Report

Transposition-Driven Genomic Heterogeneity in the Drosophila Brain

See allHide authors and affiliations

Science  05 Apr 2013:
Vol. 340, Issue 6128, pp. 91-95
DOI: 10.1126/science.1231965

Neuronal Transposons

Transposons comprise a hefty chunk of the Drosophila genome and, unregulated, can generate mutations; thus, mechanisms exist to suppress transposon activity, particularly in the germline. Perrat et al. (p. 91) investigated transposon motility in neurons of the Drosophila brain. The mushroom body of the brain, responsible for olfactory memory, contains several different types of neurons. One class of neurons, the αβ neurons, exhibited increased transposon mobility, which generated increased neuronal diversity.

Abstract

Recent studies in mammals have documented the neural expression and mobility of retrotransposons and have suggested that neural genomes are diverse mosaics. We found that transposition occurs among memory-relevant neurons in the Drosophila brain. Cell type–specific gene expression profiling revealed that transposon expression is more abundant in mushroom body (MB) αβ neurons than in neighboring MB neurons. The Piwi-interacting RNA (piRNA) proteins Aubergine and Argonaute 3, known to suppress transposons in the fly germline, are expressed in the brain and appear less abundant in αβ MB neurons. Loss of piRNA proteins correlates with elevated transposon expression in the brain. Paired-end deep sequencing identified more than 200 de novo transposon insertions in αβ neurons, including insertions into memory-relevant loci. Our observations indicate that genomic heterogeneity is a conserved feature of the brain.

Transposons constitute nearly 45% of the human genome and 15 to 20% of the fly genome (13). Mobilized transposons can act as insertional mutagens and create lesions where they once resided (2). Recombination between homologous transposons can also delete intervening loci. Specific regions of the mammalian brain, such as the hippocampus, might be particularly predisposed to transposition (4, 5). LINE-1 (L1) retrotransposons mobilized during differentiation appear to insert in the open chromatin of neurally expressed genes (46). One such insertion in neural progenitor cells altered the expression of the receiving gene and the subsequent maturation of these cells into neurons (6). The mosaic nature of transposition could therefore provide additional neural diversity that might contribute to behavioral individuality and/or neurological disorders (79).

The Drosophila melanogaster mushroom bodies (MBs) are brain structures critical for olfactory memory. The approximately 2000 intrinsic MB neurons are divisible into α′β′, γ, and αβ according to their morphology and roles in memory processing (1013). Here, we used cell type–specific gene expression profiling (14) to gain insight into cellular properties of MB neurons. Intersectional genetics (15) (Fig. 1A) allowed us to exclusively label MB α′β′, γ, and αβ neurons in the brain with green fluorescent protein (GFP) (Fig. 1, B to D). For comparison, we also assayed a “no MB” genotype in which GFP labels other neurons in the brain but not MB neurons (Fig. 1E). Sixty brains per genotype were dissected from the head capsule and dissociated by proteolysis and agitation; GFP-expressing single cell bodies were then collected by fluorescence-activated cell sorting (FACS). Total RNA was isolated from 10,000 cells per genotype, and polyadenylated RNA was amplified and hybridized to Affymetrix Drosophila 2.0 genome expression arrays. Each genotype was processed in four independent replicates (Fig. 2A).

Fig. 1

Exclusive labeling of MB neuron subsets in the fly brain. (A) MB-expressing GAL4 transgenic fly strains with expression elsewhere in the brain were intersected with a pan MB-expressing LexA. See supplementary materials for details of genetic approach. (B to E) Confocal projection of individual brains from flies, labeling α′β′ neurons (B), γ neurons (C), αβ neurons (D), and all neurons in the brain except MB neurons (E). Scale bar, 40 μm.

Fig. 2

Gene expression profiling of MB neurons. (A) Microarray data emphasizing elevated expression in αβ neurons. Each signal column per category represents an independent replicate. Fold change (FC) and t-test P values are shown for αβ versus other MB neurons (average value for α′β′ and γ) and αβ versus the no-MB group. Transposons are denoted in red. Color scale bars are linear for FC, –log10 for P value, and log2 for signal level. (B) qRT-PCR validation of increased transposon mRNA levels in αβ neurons. *αβ signal significantly different from all other populations, all P ≤ 0.05 (t test); **αβ signal significantly different from all except γ for invader3; αβ signal significantly different from other MB neurons, but not from the no-MB group, for R2.

Routine statistical analysis for differentially expressed genes, including a multiple-testing correction across all 16 data sets, did not reveal significant differences at a false discovery rate (FDR) of <0.05. We therefore used CARMAweb (16) to identify 146 mRNAs whose average signal was ≥7 in αβ neurons and that were also higher than in α′β′ neurons by a factor of ≥2. Of the top 60 transcripts from this list, 29 were significantly different from α′β′ signals (by raw P value) and represent transposons (Fig. 2A and table S1). Alignment of the corresponding values from the γ and no-MB profiles showed a similarly significant bias in transposon expression over these samples. We identified retrotransposons that transpose via a replicative mechanism involving an RNA intermediate and DNA elements that use nonreplicative excision and repair. Retrotransposons can be subdivided into long-terminal repeat (LTR) elements and long interspersed nuclear elements (LINEs). We found 11 LTR elements (Tabor, mdg1, roo, qbert, gypsy, invader3, gypsy2, microcopia, 412, accord, and blood), 11 LINE-like elements (G6, RT1b, HeT-A, Ivk, Cr1a, F element, Doc2, baggins, R2, Doc3, and Doc), and four DNA elements (Bari1, pogo, Tc3, and transib3).

We further analyzed 14 transposons, representing the most abundant in each class. Quantitative reverse transcription polymerase chain reaction (qRT-PCR) of RNA from independently purified cell samples confirmed that transposon expression was significantly higher in αβ neurons than in other MB neurons (Fig. 2B). All transposons, other than R2, were also significantly higher in αβ neurons than in the rest of the brain. R2 is unique, because it exclusively inserts in the highly repeated 28S rRNA locus and heterochromatin (17).

Transposition is ordinarily regulated by chromatin structure and posttranscriptional degradation of transposon mRNA guided by complementary RNAs (1825). The small interfering RNA (siRNA) pathway has been implicated in somatic cells (26). In contrast, the Piwi-interacting RNA (piRNA) pathway has a more established role in the germline (1921). The microarray analysis skewed our attention toward piRNA because the expression level of the translocated Stellate locus, Stellate12D orphon (Ste12DOR) mRNA, was higher in αβ than in other MB neurons and the rest of the brain by a factor of >20 (Fig. 2A and table S1). Stellate repeat transcripts are usually curtailed by piRNA, not siRNA (27, 28). Stellate repeats encode a casein kinase II regulatory subunit, and piRNA mutant flies form Stellate protein crystals in testis (29). Immunostaining Stellate in the brain labeled puncta within αβ dendrites in the MB calyx, consistent with high Ste12DOR expression in wild-type αβ neurons (fig. S1).

piRNAs are loaded into the Piwi clade argonaute proteins Piwi, Aubergine (Aub), and Argonaute 3 (Ago3) (21). Piwi and Aub can amplify piRNA pools with Ago3 (30, 31). To investigate piRNA involvement in differential transposon expression, we immunolocalized Piwi proteins and colocalized GFP to assign signals to MB neuron type (Fig. 3). Aub and Ago3 differentially labeled MB subdivisions in addition to structures throughout the brain (Fig. 3 and figs. S2 and S3), but we did not detect Piwi (32). The ellipsoid body of the central complex stained strongly for Aub but not at all for Ago3 (Fig. 3, A and E, and figs. S2D and S3D), which suggests possible functional exclusivity of Piwi proteins in the brain.

Fig. 3

Aub and Ago3 are not abundant in αβ neurons. (A) Aub immunostaining (magenta) labels the ellipsoid body (EB), MB subdivision in the peduncle (ped), and cell body layer (rind). A single confocal section is shown at the level of the MB peduncle. Dashed box denotes area shown in (B) to (D). (B to D) Aub labels α′β′ and γ neurons in the peduncle more than αβ neurons. Aub staining (B) is mutually exclusive to GFP-expressing αβ neurons (green) but overlaps with α′β′ neurons (C) and γ neurons (D). (E) Ago3 immunostaining (magenta) labels neurons throughout the brain and MB subdivision in the peduncle. A single confocal section is shown at the level of the MB peduncle. Dashed box denotes area shown in (F) to (H). (F to H) Ago3 staining is prominent in the αβ core [(F), solid triangle] and does not overlap with outer GFP-labeled αβ neurons [(F), open triangle], nor with α′β′ neurons (G), but overlaps with γ neurons (H). Scale bars, 40 μm [(A) and (E)], 10 μm [(B) to (D), (F) to (H)]. (I) qRT-PCR analysis from wild-type, aubHN2/aubQC42, ago3t2/ago3t3, and armi1/armi27.1 mutant heads. Several transposon transcripts are elevated in ago3, aub, and armi mutant fly brains. Values are normalized to wild-type heads. *Significant increase, P < 0.05 (t test).

Differential Aub and Ago3 labeling was most evident within axon bundles in the peduncle and lobes, where MB neuron types are anatomically discrete (Fig. 3, A to H, and fig. S2). Aub protein colocalized with γ and α′β′ neurons in the peduncle and lobes but was reduced in αβ neurons in both locations (Fig. 3, B to D, and fig. S2). Ago3 did not label MB lobes (fig. S3) but colocalized with γ neurons in the peduncle (Fig. 3H). Ago3 labeled core αβ (αβc) neurons but did not label outer αβ neurons (Fig. 3F). Therefore, outer αβ neurons do not abundantly express Aub or Ago3, which implies that transposon suppression is relaxed. In contrast, γ neurons express Aub and Ago3, providing potential for piRNA amplification, and α′β′ neurons express Aub. These patterns of Aub and Ago3 in the MB peduncle appear conserved in brains from D. erecta, D. sechellia, and the more distantly related D. pseudoobscura species (fig. S4).

Loss of siRNA function elevates transposon expression in the head (22). We replicated these findings with ago2414 and dcr-2L811fsX mutant flies (fig. S5). In parallel, we used trans-heterozygous aub (aubHN2/aubQC42) and ago3 heads (ago3t2/ago3t3) (19, 33) and trans-heterozygous armitage heads (armi1/armi72.1) (34) to test whether piRNA suppressed transposon expression (Fig. 3I). Levels of the 14 LTR, LINE-like, and TIR group transposons verified to be expressed in αβ neurons (Fig. 2B) were assayed by qRT-PCR; of these 14 transposons, 13 were significantly elevated in siRNA-defective ago2 and dcr-2 mutants (fig. S5). The piRNA-defective aub, ago3, and armi mutants also exhibited significantly elevated levels of 9 of the 14 elements (Fig. 3I). Levels of the LTR elements gypsy, Tabor, and qbert; the LINE-like elements HeT-A, RT1b, and R2; and the TIR element pogo were higher in ago3 mutants. In addition, blood, Tabor, and R2 were elevated in aub mutants, and blood, gypsy, Tabor, invader3, qbert, HeT-A, and R2 were elevated in armi mutants. Therefore, the piRNA pathway contributes to transposon silencing in the brain, and low levels of Aub and Ago3 may permit expression in αβ neurons.

To determine whether transposons are mobile, we mapped new insertions by deep sequencing (35) of αβ DNA (Fig. 4). We purified αβ neurons by FACS, as for transcriptome analysis, but isolated genomic DNA. Insertions were defined by paired-end reads in which one end mapped to the annotated genome and the other to the transposon sequence. To identify de novo transposition events in αβ neurons, we compared the genomic position of transposons within the αβ sequence to those located by sequencing DNA from genetically identical embryos. In addition, we sequenced DNA from the remainder of the brain tissue from the FACS separation of αβ neurons (Fig. 4A and table S4).

Fig. 4

Identification of de novo transposition events in αβ neurons. (A) Venn diagram of transposon insertions identified in this study. Direct genome sequencing identified 215 insertion sites in αβ neuron DNA and 200 in DNA from other brain cells that were not present in embryo DNA from genetically identical sibling flies. (B) Chromosomal distribution of new transposon insertions in αβ neurons. Bar height indicates number of insertions in each location. piRNA clusters are shown. (C) The proportion of new insertions found for each transposon class in αβ neurons. (D) Proportion of insertions per transposon class in the inherited fly genome from our sequencing of embryo DNA. (E) Distribution of transposons in the annotated genome significantly differs from distribution of de novo insertions in αβ neurons, the rest of the brain, and ovary DNA with respect to neighboring genes (P < 2.2 × 10–16 for αβ neurons, the rest of the brain, and ovary; χ2 test).

These studies identified 3890 transposon insertions in embryo DNA that differed from the published Drosophila genome sequence (Fig. 4A). In comparison, αβ neuron DNA revealed 215 additional sites. The remaining brain tissue uncovered 200 new insertions, including 19 that were identical to those in αβ neurons. The sequencing depth for embryos (34.1×) was an order of magnitude greater than for neurons (3.1×) because embryo material could be collected more easily; hence, the αβ and other brain insertions are likely de novo. By randomly sampling reads to yield 1× genome coverage, we calculated 129 new transposon insertions per αβ neuron genome (table S2). Sequencing single neurons (36) would reveal the exact cellular frequency and heterogeneity of transposition events.

New αβ insertions occurred across all chromosomes, without obvious regional bias (Fig. 4B). In addition, insertions resulted from 49 different transposons representing LTR, LINE-like, TIR, and Foldback (FB) classes (Fig. 4C). They included 11 of the 29 transposons in the αβ transcriptome (Fig. 2A), and the number of insertions per class was consistent with their prevalence in the genome (Fig. 4D). Therefore, many transposons mobilize in αβ neurons.

Of the 215 de novo αβ insertions, 108 mapped close to identified genes (Fig. 4E and fig. S6). Of these, 35 disrupted exons, 68 disrupted introns, and 5 fell in promoter regions (<1 kb from transcription start site). The remaining 107 insertions mapped to piRNA clusters or intergenic regions and were not assigned to a particular gene. A similar distribution was observed for the 200 new insertions in the rest of the brain (Fig. 4E and fig. S6). The reference fly genome has 258 transposon insertions in exonic regions, 11,110 insertions in intronic regions, 502 insertions in promoter regions, and 33,008 insertions in intergenic regions. Therefore, both groups of brain cells had a significantly larger fraction of insertions within exons, and fewer in intergenic regions, than the transposons that are annotated in the genome (Fig. 4E and table S8). To test whether such a distribution was unique to neurons, we analyzed de novo insertions in ovary DNA, again using embryo sequence as the comparison. New insertions in ovary DNA revealed a similar skew toward exons (Fig. 4E and table S8).

In mammals, active L1 elements appear to disrupt neurally expressed genes (46). New αβ neuron insertions, but not those in other tissue (tables S7 and S9), were significantly enriched in 12 Gene Ontology (GO) terms (Benjamini FDR <5%; tables S4 and S5), all of which are related to neural functions. Moreover, promoter regions from 18 of 20 of the targeted genes drive expression in αβ neurons (table S10). We found exonic insertions in gilgamesh, derailed, and mushroom body defect and intronic insertions in dunce and rutabaga (table S3), all of which have established roles in MB development and function (3740). In addition, MB neurons are principally driven by cholinergic olfactory projection neurons (41) and receive broad GABA-ergic inhibition (42) and dopaminergic modulation through G protein–coupled receptors (43). We identified intronic insertions in nicotinic Acetylcholine Receptor α 80B, G protein-coupled receptor kinase 1, and cyclic nucleotide gated channel-like and an exonic insertion in GABA-B-receptor subtype 1 (table S3). Transposon-induced mosaicism could therefore alter integrative and plastic properties of individual MB αβ neurons.

Our data establish that transposon-mediated genomic heterogeneity is a feature of the fly brain and possibly other tissues. Together with prior work in rodents and humans (46), our results suggest that genetic mosaicism may be a conserved characteristic of certain neurons. Work in mammals indicates that L1 expression occurs because the L1 promoter is released during neurogenesis (6, 21). Our data are consistent with such a model and also support the idea that transposons avoid posttranscriptional piRNA silencing in adult αβ neurons.

A recent study described a role for piRNA in epigenetic control of memory-related gene expression in Aplysia neurons (44). It is therefore possible that MB neurons differentially use piRNA to control memory-relevant gene expression and that transposon mobilization is an associated cost. Because we found transposon expression in αβ neurons of adult flies, it is conceivable that disruptive insertions accumulate throughout life, leading to neural decline and cognitive dysfunction. Alternatively, permitting transposition may confer unique properties across the 1000 neurons in the αβ ensemble and potentially produce behavioral variability between individual flies in the population.

Supplementary Materials

www.sciencemag.org/cgi/content/full/340/6128/91/DC1

Materials and Methods

Figs. S1 to S6

Tables S1 to S10

References (4549)

References and Notes

  1. Acknowledgments: Data are available through ArrayExpress (www.ebi.ac.uk/arrayexpress) and NCBI Short Read (www.ncbi.nlm.nih.gov/sites/sra) archives as E-MEXP-3798 and SRP017718, respectively. We thank A. Vodala, J. Menet, K. Abruzzi, and N. Francis for advice with techniques. Supported by a Wellcome Trust Senior Research Fellowship in Basic Biomedical Sciences, the Gatsby Charitable Foundation, Oxford Martin School and grants MH069883 and MH081982 from NlH (S.W.), NIH grants NS044232 and NS045713 and the Ellison Medical Foundation (M.R.), and NIH grant R01HD049116 (Z.W. and W.T.).
View Abstract

Navigate This Article