(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Capsid transfer of the retrotransposon Copia controls structural synaptic plasticity in Drosophila [1]

['P. Githure M Angale', 'Department Of Neurobiology', 'University Of Massachusetts Chan Medical School', 'Worcester', 'Massachusetts', 'United States Of America', 'Adrienne Lemieux', 'Yumeng Liu', 'Department Of Biochemistry', 'Molecular Biotechnology']

Date: 2025-02 

Transposons are parasitic genome elements that can also serve as raw material for the evolution of new cellular functions. However, how retrotransposons are selected and domesticated by host organisms to modulate synaptic plasticity remains largely unknown. Here, we show that the Ty1 retrotransposon Copia forms virus-like capsids in vivo and transfers between cells. Copia is enriched at the Drosophila neuromuscular junction (NMJ) and transported across synapses, and disrupting its expression promotes both synapse development and structural synaptic plasticity. We show that proper synaptic plasticity is maintained in Drosophila by the balance of Copia and the Arc1 (activity-regulated cytoskeleton-associated protein) homolog. High-resolution cryogenic-electron microscopy imaging shows that the structure of the Copia capsid has a large capacity and pores like retroviruses but is distinct from domesticated capsids such as dArc1. Our results suggest a fully functional transposon mediates synaptic plasticity, possibly representing an early stage of domestication of a retrotransposon.

We found Copia can form capsids in a cell-free manner and that it associates with its own mRNA in vivo, suggesting that Copia retains its viral-like characteristics of trafficking its RNA. We developed a suite of tools, including antibodies and tissue-specific manipulation of Copia expression, and found that the capsid region of Copia is enriched at and transfers across Drosophila NMJs. Strikingly, disrupting Copia expression at the NMJ led to an increase in synapse formation and structural synaptic plasticity. We further found that Copia and dArc1 have an antagonistic relationship phenotypically and genetically. Because Copia behaves in a viral-like manner, we used cryogenic-electron microscopy (cryo-EM) to determine the high-resolution structure of the Copia capsid and found that it is distinct from the related dArc1 capsid, with a larger internal capacity and the presence of pores similar to those used by retroviruses to carry out their reverse transcriptase activity. The Copia capsid also contains a large amount of RNA, which we show is necessary for its assembly. Taken together, our observations suggest that Copia is a critical factor in NMJ plasticity that may interact with a well-established plasticity pathway. This novel form of TE domestication provides some of the first in vivo evidence that TEs play a pivotal role in neuronal development.

The remarkable viral-like mechanism of Arc compelled us to examine how other capsid-forming elements may contribute to neuronal development and synaptic plasticity. We had previously found that RNA transcripts of Copia, an active retrotransposon, are also enriched in extracellular vesicles (EVs) at the NMJ [ 5 ]. Copia is known to form capsid-like structures [ 8 – 12 ], which prompted us to explore its relationship to neuronal development.

Recent work has revealed that the capsid regions of RTEs have been domesticated numerous times throughout Metazoan evolution. For example, we and others discovered that the gene Arc (activity regulated cytoskeleton-associated; dArc1 in Drosophila) is derived from a retroviral capsid region and retains the ability to form capsids [ 5 , 6 ]. These capsids carry RNA in extracellular vesicles across the neuromuscular junction (NMJ), which is important for synaptic plasticity and development. Thus, Arc has retained its viral-like behavior of RNA transfer to mediate its function in establishing synaptic plasticity. However, the dArc1 capsid structure is much smaller and less porous than retroviral counterparts [ 7 ], suggesting that dArc1’s mechanism has been specialized following its domestication.

Retrotransposons (RTEs) are robust sources of domesticated genes, especially in mammals [ 3 ]. They are selfish genetic elements that replicate using an RNA intermediate which becomes reverse transcribed into DNA for integration at another site. Thus, RTEs use a “copy and paste” strategy for replication to move and spread throughout the host genome. Approximately 40% of the human genome consists of RTEs, illustrating the effectiveness of their reproductive strategy though only a small subset of these elements remains active in the genome. The long terminal repeat retrotransposon replication mechanism is similar to that of retroviruses such as human immunodeficiency virus (HIV) and uses related proteins. For example, similar GAG proteins form capsid shells that protect and transport the RNA intermediate, and related enzymes such as reverse transcriptase and integrase are used to replicate the sequence and insert it into the genome. In fact, retrotransposons (RTEs) are endogenous retroviruses of the viral genus Metaviridae [ 4 ].

Transposable elements (TEs) are DNA sequences that can move within the genome. TEs were once thought of as being “junk DNA,” and in ALS and Alzheimer’s Drosophila models suggest that TEs may be contributing to the pathologies of neurodegenerative diseases [ 1 , 2 ]. Despite these deleterious roles, there is considerable evidence that TEs provide genetic variability for the evolution of new molecular, cellular, and organismal functions. When a TE-derived DNA segment is selected for, its fitness benefits the host, and it can become a “domesticated” gene. Though it remains unclear how TE domestication occurs, particularly at the early stages.

The number of cycles of spaced K + stimulation required for ghost bouton formation can be used as a measure of acute structural synaptic plasticity [ 29 ]. Therefore, we sought to assess whether Copia knockdown altered the threshold for K + stimulated formation of ghost boutons. As expected for wild-type Drosophila, 3 cycles of spaced K + stimulation are insufficient to initiate ghost bouton formation, while 5 cycles induce robust ghost bouton formation ( Fig 5F–5I ). However, disrupting Copia expression presynaptically induced significant ghost bouton formation with only 3 cycles of K + stimulation ( Fig 5F–5I ). This result shows that presynaptic reduction of Copia increases acute plasticity.

Our discovery that Copia is a negative regulator of bouton formation, which is an indication of developmental plasticity, prompted us to test if Copia has acute roles in regulating plasticity. New synaptic bouton formation can be stimulated in dissected larval NMJs with acutely repetitive spaced cycles of nerve stimulation with potassium (High K + ; 90 mM) [ 29 ]. These nascent boutons, however, do not properly develop pre- and postsynaptic structures, and are thus called ghost boutons. In previous work, we found that NMJs from dArc1-null flies were less responsive to spaced K + stimulation, which resulted in decreased formation of ghost boutons compared to controls [ 5 ]. Since larvae expressing Copia pol -shRNA and Copia gag -siRNA already have a larger number of synaptic boutons, Copia knockdowns have increased structural synaptic plasticity.

As dArc1 and Copia have an inverse relationship such that reduction of one at the NMJ leads to an increase of the other, we assessed if dArc1 and Copia genetically interact to determine their functional relationship. We disrupted Copia expression using either Copia pol -shRNA or Copia gag -siRNA in a dArc1-null mutant background. There was a substantial decrease in bouton formation in dArc1-null mutants, which was consistent with our previous studies [ 5 ] ( S6F Fig ). Reduction of Copia in neurons using either Copia pol -shRNA or Copia gag -siRNA in the dArc1-null background resulted in a substantial increase in bouton numbers and hyperbudding ( S6A–S6G Fig ), similar to levels observed with Copia pol -shRNA or Copia gag -siRNA expression in a wild-type background. These results are consistent with Copia having a predominant role at the NMJ over dArc1 ( S6H–S6K Fig ).

(A–C) Expressing Copia pol -shRNA (B) or Copia gag -siRNA (C) constructs in the motoneurons of larvae causes increased bouton formation at NMJs compared to wild type (C380-Gal4/Canton-S) (A). There is also a substantial increase in hyperbudding (A3 vs. B3 and C3), where 2 or more boutons “bud” off the same central bouton (yellow arrows designate budding boutons in B3 and C3). (D) Quantification of bouton numbers. (E) Quantification of hyperbudding events. (F–H) Stimulating wild-type flies (F) with 3 rounds of potassium treatments does not induce them to form new synapses (ghost boutons) but does induce increased bouton formation after disrupting expression of Copia (G and H). (I) Quantification of potassium stimulation in different genetic backgrounds. The data underlying the graphs shown in the figure can be found in S1 Data . DLG = α-Discs Large (postsynaptic marker), HRP = α-horseradish peroxidase (presynaptic marker). N for D and E = from left to right; number of animals/NMJs quantified, 14/27, 9/17, 9/16, for I N = 12/19, 10/13, 6/12, 9/17, 6/11, 9/18. WT is C380-Gal4/Canton-S for panels A and F. ns p ≥ 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001. NMJ, neuromuscular junction.

Because Copia gag is enriched at the NMJ, we investigated its potential role in synaptic development and plasticity. Synaptic boutons at the NMJ are continuously formed throughout larval development [ 27 ]. Thus, an increase or decrease in the levels of bouton formation is an indication of developmental synaptic plasticity [ 28 ]. When we express Copia pol -shRNA with the neuronal C380-Gal4 driver to specifically reduce Copia expression presynaptically, we observe a striking ~50% increase in synaptic bouton number compared to the C380-Gal4 driver alone control ( Fig 5A–5D ). This effect was also seen with the single siRNA construct directed against the Copia gag mRNA splice site ( Fig 5A, 5C, and 5D ). In addition to an enhancement in the number of synaptic boutons upon presynaptic expression of either Copia pol -shRNA or Copia gag -siRNA, we also observed an increase in “hyperbudding,” which we defined as the presence of 3 or more boutons budding off from a central, larger (parent) bouton ( Fig 5E ). Thus, Copia negatively regulates developmental plasticity, while Arc is a positive regulator. Consistent with this, overexpression of dArc1 at the NMJ increased the number of boutons, which shows that elevated levels of dArc1 can overcome Copia-induced inhibition of bouton formation ( S6F Fig ).

Because the Copia capsid structure exhibits less extensive interactions than other related capsids, we sought to determine the requirements for Copia capsid assembly. We found that Copia capsid formation requires the presence of RNA, as removing bound RNA using anion-exchange chromatography inhibited capsid assembly ( S3C and S3D Fig ). Likewise, deleting the RNA-binding Zn-finger domain (i.e., the Copia gag -ΔNC variant) inhibited capsid assembly ( S3E Fig ). However, the Copia gag -ΔNC variant was capable of co-assembly into capsid-like particles in the presence of WT-Copia gag carrying RNA ( S3F Fig ). Thus, the Copia gag -ΔNC variant is assembly-competent as long as RNA-binding is achieved by some of the co-assembling subunits. These results establish a requirement for RNA in normal Copia capsid assembly. This is distinct from dArc1 capsids that can assemble in the absence of RNA [ 7 ], or mammalian Arc which assembles in absence of RNA albeit much less efficiently than with RNA bound [ 6 ]. Copia’s RNA requirement is much more similar to HIV, which requires RNA binding for assembly [ 26 ]. These results further demonstrate that Copia is functioning in a virus-like/retrotransposon fashion.

Copia forms less interactions within a capsomer than other retroelement-derived capsids for which high-resolution structures are available ( S5B Fig ). For example, Copia capsomers have ~20% less interaction area than HIV capsomers [ 20 ]. The Copia interaction area is ~5% less than the modeled regions of the dArc1 capsomers, even though the dArc1 spike protrusions that constitute a massive inter-subunit interaction are unable to be modeled in the dArc1 capsid structure [ 7 ]. Despite the less extensive interfaces between capsid subunits in Copia, we observed an interesting interaction between adjacent CTDs that is not found in other retroelement-derived capsids ( S5B Fig ). We suspect this CTD-CTD interaction plays a key role in stabilizing Copia capsids by locking down the periphery of each capsomer sub-assembly. The Ty3/Gypsy capsid also has this CTD-CTD interaction [ 25 ], suggesting that it is shared among the Ty1 and Ty3 families of retrotransposons ( S5B Fig ).

The Copia capsid is quite different than that of Arc. The Copia capsid is much larger than that of the T = 4 dArc1 capsid (inner radii of ~220 Å and ~170 Å, respectively), resulting in ~2.2× increased internal capacity ( Fig 4C ). We hypothesize that the Copia capsid is larger than dArc1 because it packages larger cargo, such as the longer Copia mRNA (5 kbp versus 2 kbp) as well as the Copia-encoded enzymes reverse transcriptase and/or integrase. In addition to differences in size, Copia and dArc1 capsids are formed using quite distinct types of interactions, with very dissimilar electrostatic surface potentials ( S5A Fig ). This implies that Copia and dArc1 capsid proteins cannot physically interact with each other. Moreover, the Copia capsid also lacks the spike structures that protrude from the outer surface of the dArc1 capsid [ 7 ]. In place of the spikes, the Copia capsid contains positively charged pores at the center of each hexameric or pentameric sub-assembly, similar to HIV [ 20 ] ( Fig 4D ). The positively charged pores of HIV function as gates for dNTP entry into the capsid for fueling reverse transcriptase activity [ 21 ]. We hypothesize that the pores in the Copia capsid play a similar role. In support of this hypothesis, the capsids encoded by the domesticated GAG genes (dArc1 and PNMA2) both display spikes that block these pores, presumably because dArc1 and PNMA2 capsids do not harbor an active reverse transcriptase [ 22 ]. In contrast, structures of active retrovirus or retrotransposon capsids exhibit open pores [ 20 , 23 – 25 ] presumably to facilitate reverse transcription.

The capsid layer is composed of the Copia gag protein’s CA region, which contains 2 alpha-helical domains ( Fig 4B ). The CA region is attached through a flexible linker to the nucleocapsid (NC) region, which harbors a Zn-finger RNA-binding domain. The NC region is not clearly visualized in the map but is located in the interior of the capsid and is associated with the density attributed to RNA. The N- and C-terminal domains (NTD and CTD) of the CA region are alpha-helical domains that are joined by a flexible linker. The NTDs form central spokes at the center of each capsomer sub-assembly, while the CTD is arranged in the capsomer periphery. The NTD and CTD of Copia are structurally similar to those seen in retroviruses and other retrovirus-like capsids, such as dArc1 (C a RMSD ~3.5 Å and 2.7 Å for the NTD and CTD, respectively) [ 7 ].

( A) The electron density map of the Copia Gag capsid, colored by radius. The left half of the map is the surface of the capsid, and the right half of the map is a cross-section view. The 5-fold pentamer, 3-fold hexamer, and non-symmetric hexamer are denoted as black pentagon, triangle, and hexagons, respectively. The asymmetric unit is outlined in a magenta dashed line on the map and shown in detail on the top left panel, highlighting the 9 monomers composing one asymmetric unit (T = 9). The radius (220 Å) is measured between the center of the capsid and the mass center of the Copia CA_CTD of the asymmetric unit. The outer layer is labeled as capsid (CA) layer, and the yellow inner layer density is labeled as nucleocapsid (NC) and RNA layer, the structure of which is undetermined due to disorder. (B) The atomic model of the Copia hexamer. Top panel: model of the CA region is colored with NTD in dark gray and CTD in light gray. The unstructured NC domain is shown as a black line with the zinc finger domain (ZnF) located near the C-terminus. One monomer within the hexamer is colored in rainbow from the N- to C- terminus. Bottom panel: the atomic model of one Copia Gag monomer. (C) The electron density map of the dArc1 capsid colored by radius and was determined in a previous study [ 7 ]. The left half of the map is the surface of the capsid; the right half of the map is a cross-section view. The radius (170 Å) is measured between the center of the capsid and the mass center of the dArc1 CA_CTD . The outer layer is labeled as capsid (CA) layer; the spikes on the surface of dArc1 capsid are unmodeled. There is no bound RNA in the assembled dArc1 capsid. (D) Copia Gag non-symmetric hexamer is colored by electrostatic potential. The positively charged (blue) pore is at the center of the hexamer. The data underlying the structure shown in the figure can be found in S1 Data . CTD, C-terminal domain; NTD, N-terminal domain.

Our purification protocol enriched capsids of sufficient homogeneity such that we could determine their structure using single-particle cryo-EM methods ( S4A Fig ). Our reconstruction of the Copia gag capsid when enforcing T = 9 icosahedral symmetry reached ~4.2-Å overall resolution (Figs 4A and S4B ). There is a layer of featureless density lining the inner surface of the Copia capsid that likely represents the RNA that comigrates with the Copia-gag protein. Because the RNA does not adopt T = 9 symmetry, its density is amorphous in our reconstruction. To enhance the resolution of the capsid region, we used local reconstruction methods to determine the structures of individual subregions of capsid known as capsomers, which are sub-assemblies that are found at the vertices or faces of the icosahedron ( S4D Fig ). Local reconstruction of the pentameric capsomers found at the vertices provided a map to an overall resolution of ~3.3-Å. Because the Copia capsid adopts T = 9 geometry, there are 2 different hexameric capsomers along the faces (a hexamer with local 3-fold symmetry and a non-symmetric hexamer) and local refinement of each hexamer provided high-resolution maps (overall resolution of the non-symmetric and 3-fold symmetric hexamers are ~3.5-Å and 3.4-Å overall resolution, respectively). The high-quality maps of the locally refined capsomers allowed us to build an atomic model of the Copia gag protein with high confidence and good stereochemistry (Figs 4B and S4C ).

Both dArc1 and Copia are retroviral-like factors that form capsids and have mutually antagonizing effects at the larval NMJ. To further investigate how Copia behaves in a virus-like manner, we sought to determine the three-dimensional structure of the Copia capsid. We expressed recombinant His6-SUMO-Copia gag protein in E. coli and purified the protein to homogeneity ( S3A–S3C Fig ). We observed efficient auto-processing of the Copia gag protein within the E. coli to form separate capsid (CA) and protease (PR) proteins, indicating that auto-processing does not require any Drosophila-encoded factors ( S3A and S3B Fig ). This result is similar to an earlier report showing auto-processing of full-length Copia polyprotein in E. coli [ 16 ]. Upon removal of the N-terminal His6-SUMO domain, we observed self-assembly of the Copia gag protein into capsid-like particles of assorted sizes as well as some irregular aggregates.

(A–C) The presynaptic knockdown of Copia with Copia pol -shRNA (B1–B4) and Copia gag -siRNA (C1–C4) leads to a significant increase in α-dArc1 staining both pre- and postsynaptically at the larval NMJ compared to wild type (A1–A4). (D, E) The presynaptic knockdown of dArc1 (E1–E4) leads to a significant increase in α-Copia gag staining both pre- and postsynaptically at the larval NMJ compared to wild type (D1–D4). (F–I) Quantification of the data represented in A–E. Quantification of dArc1 mRNA in the larval CNS (J), and BWM (K) of strains used in this study, showing a significant decrease of copia mRNA in lines with both the Gal4 driver and copia si/shRNA. The data underlying the graphs shown in the figure can be found in S1 Data . DLG = α-Discs Large (postsynaptic marker), HRP = α-horseradish peroxidase (presynaptic marker). N = number of animals/NMJs quantified in (A) 8/14, (B) 8/14, (C) 8/14, (D) 11/20, and (E) 8/13. WT is C380-Gal4/Canton-S for panels A, D, and F–I. ns p ≥ 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001. BWM, body wall muscle; CNS, central nervous system; NMJ, neuromuscular junction.

Copia and dArc1 both form capsids that bind to their RNAs and are transferred across the NMJ through EVs. To test if the 2 proteins act antagonistically, we examined whether Copia levels affect dArc1 levels and vice versa. Presynaptic disruption of Copia using Copia pol -shRNA or Copia gag -siRNA led to a large accumulation of α-dArc1 signal both pre- and postsynaptically ( Fig 3A–3C , 3F , and 3G ). Additionally, presynaptic knockdown of dArc1 in flies resulted in an increase of α-Copia gag signal in the pre- and postsynaptic compartments ( Fig 3D , 3E , 3H , and 3I ). Thus, reduction of either Copia or dArc1 results in increased protein levels of the opposing factor. Consistent with an increase of dArc1 protein at the NMJ in larvae expressing Copia-RNAi constructs, we observed an increase of dArc1 mRNA in the pre- and postsynaptic tissue as determined by dPCR ( Fig 3J and 3K ). This suggests that the regulation of these opposing factors occurs at the mRNA level.

We next asked whether Copia or dArc1 interact with each other’s mRNA. We had previously observed that dArc1 protein does not precipitate Copia mRNA [ 5 ]. Consistent with this, sequencing of bound RNA following immunoprecipitation (RIP-Seq) using the α-dArc1 antibody did not show enrichment for Copia mRNA ( Fig 2H ). In contrast, Copia immunoprecipitated with both Copia mRNA and dArc1 mRNA ( Fig 2I ). We confirmed the association of Copia gag with both Copia and dArc1 transcripts using dPCR ( S2K–S2P Fig ). Thus, our results indicate that dArc1 and Copia form separate capsid structures and traffic through separate EVs, but Copia capsids bind both Copia and dArc1 mRNA while dArc1 only associates with its own transcript.

We next investigated whether Copia and dArc1 capsids co-localize in EVs. We isolated EVs from S2 cells, exposed them to saponin, to permeabilize the EV membranes, but keep EVs intact. We then incubated these saponin-treated EVs with α-Copia gag and α-dArc1 and imaged using immuno-EM. We found that there is a striking exclusion of Copia from dArc1-labeled EVs and of dArc1 from Copia-labelled EVs ( Fig 2E and 2F ) . We also double labeled some saponin treated EVs with Copiagag and Syntaxin-1A, a well-characterized EV marker in Drosophila and found that Copia and Syntaxin-1A co-labeled some of the EVs ( S2G Fig ). We further performed this experiment with harsher detergent treatments that remove the EV membranes (as in Fig 2C and 2D , quantified in Figs 2G , S2H , and S2J ) and found that the capsid-like electron dense structures only label with either α-dArc1 or α-Copia gag . We performed secondary antibody controls and found no immunoreactivity ( S2I Fig ). We also found that electron densities that were labelled with Copia antibody were about 55 to 60 nm ( S2J Fig ) which is consistent with the size we observe as the predominant species of in vitro formed EVs (Figs 2B and S1J ).

We previously found that Copia gag mRNA was enriched in EVs derived from Drosophila S2 cultured cells [ 5 ]. To determine whether Copia gag capsids are also present in EVs, we isolated EVs from S2 cells, lysed them with detergent, and immunostained with α-Copia gag then observed using EM. We observed electron densities consistent with capsids labelled with the Copia antibody, suggesting that these EVs contain Copia capsids ( Fig 2C and 2D , quantified in 2G). Capsid proteins surround and protect the viral genome during its transfer between cells. Thus, we tested whether Copia gag associates with Copia transcripts in vivo by conducting RNA immunoprecipitation (RIP) experiments using larval CNS and BWM extracts dissected from wild-type Drosophila third-instar larvae. We found that both α-Copia Full and α-Copia gag immunoprecipitated the Copia transcript from both of these compartments in vivo ( S2A–S2F Fig ). This data indicates that Copia forms capsids in vivo and encapsulates its own transcript.

(A) Bacterially expressed Copia gag self-assembles into capsid-like structures, which are observable using negative stain EM, scale bar = 200 nm. (B) Close-up of an individual capsid, scale bar = 50 nm. (C) Representative EVs isolated from S2 cells were immuno-stained with α-Copia gag (18 nm gold secondary). (D) A higher magnification of the closed square in D showing electron density of the α-Copia gag stained S2 EVs (scale bar = 50 nm). (E, F) Representative EVs isolated from S2 cells were simultaneously immuno-stained with α-Copia gag (18 nm gold secondary) (G), α-dArc1 (10nm), (H) scale bar = 50 nm. (G) Quantification of dual immuno-staining of 20 grids (representative images in 2C and 2D) were co-labelled with both α-Copia gag and α-dArc1. Particles labelled with either α-Copia gag or α-dArc1, both together, and unlabeled were counted (n = 1,042). (H, I) Mapping of RIP-seq, antibody used for immunoprecipitation on the y-axis, the genomic region of A is Copia and of B is dArc1. The data underlying the graphs shown in the figure can be found in S1 Data . EV, extracellular vesicle; RIP, RNA immunoprecipitation.

The spliced form of Copia (Copia gag ) contains the GAG region that encodes the capsid protein and a small part of the POL region that consists primarily of the protease region necessary for auto-processing and viral maturation [ 14 , 15 ] ( Fig 1A ). Consistent with previous studies, we found that bacterially expressed Copia gag auto-assembled into capsid-like structures [ 8 – 10 , 12 , 19 ] ( Fig 2A and 2B ). Copia gag protein can form capsids of varying sizes, ranging from 30 nm to more than 120 nm in diameter ( S1I and S1J Fig ). The variable sizes of these capsids have important implications for capsid formation and entry into cells and/or nuclei.

We sought to address whether our RNAi constructs exhibit off-target effects. TEs can co-splice with nearby genes, so it is possible that a TE-gene chimera could be repressed by one of our RNAi constructs [ 18 ]. We used whole genome sequencing to identify any TE-gene chimeras occurring in the Drosophila lines carrying our Gal4-UAS constructs. While we did identify rare chimeric sequences, our RNAseq data does not show any repression of the chimeric genes by our RNAi constructs (see Methods ). Thus, we attribute the effects of our Gal4-UAS RNAi constructs specifically to Copia repression.

To determine whether Copia gag is transferring across the synapse, we utilized a siRNA construct, UAS-Copia gag -siRNA, that recognizes the Copia gag splice site ( Fig 1A ). The Copia gag -siRNA is predicted to specifically disrupt Copia gag and not the full length unspliced isoform of Copia. We observed a substantial reduction of α-Copia gag signal at the NMJ when expressing Copia gag -siRNA presynaptically ( Fig 1C, 1E and 1G ). We expressed both Copia gag -siRNA and the Copia pol -shRNA in the postsynaptic muscle using the C57-Gal4 expression driver and observed a decrease of α-Copia gag signal in the postsynaptic area, but we did not see a decrease of α-Copia gag signal presynaptically ( S1D–S1H Fig ). This result suggests that the transfer of Copia gag is pre- to postsynaptic, but Copia mRNA can be affected by postsynaptic RNAi expression. Altogether this suggests that the postsynaptic localization of Copia gag is at least in part due to a Copia mRNA pool derived from the presynaptic cell. We validated that this pattern was consistent at the mRNA level, as only larvae expressing a Copia-si/shRNA had a significant decrease in Copia mRNA expression ( Fig 1H and 1I ).

We next asked whether Copia capsid protein is transported across the NMJ. Immunolabeling with α-Copia gag revealed a striking enrichment at the Drosophila NMJ ( Fig 1C ). To determine whether the Copia capsid protein is crossing the synapse, we leveraged the powerful Gal4/UAS system [ 17 ], which has been extensively used for precise genetic manipulation in pre- and/or postsynaptic cells at the NMJ. We designed 3 UAS-shRNA constructs targeting different regions of Copia ( Fig 1A ). We observed a substantial reduction of the α-Copia gag signal when expressing Copia pol -shRNA specifically in motor neurons using the presynaptic C380-Gal4 expression driver ( Fig 1D and 1F ). Notably, we see reduced α-Copia gag staining in the postsynaptic area abutting the synapses ( Fig 1G ). (See S1C Fig for a definition of the anatomy of a bouton, with pre- and postsynaptic areas highlighted.) This result strongly suggests that the postsynaptic localization of Copia gag is at least in part due to a pool of Copia RNA that is derived from the presynaptic cell.

(A) Representation of the entire Copia genome (Copia full ) and the spliced form (Copia gag ). The location of the RNAi targets, splice site (dotted lines), where the spliced ends are joined (thick black arrow) and the peptide used to generate the Copia gag antibody are indicated. (B) Western blot of larval CNS and BWM lysate with the predicted size of Copia gag indicated on the right and a molecular weight ladder on the left. (C) Representative confocal images of α-Copia gag immunostaining show a striking enrichment at the wild-type Drosophila larval NMJ. (D, E) Knockdown with either Copia pol -shRNA or Copia gag -siRNA constructs expressed in the presynaptic compartment (C380-Gal4) cause a clear reduction of α-Copia gag in both pre- and postsynaptic sides of the NMJ compared to wild type (C380-Gal4/Canton-S). (F) Quantification of presynaptic α-Copia gag normalized immunofluorescence and postsynaptic (G) intensity of the same. Quantification of copia mRNA in the larval CNS (H) and BWM (I) by digital PCR of strains used in this study showing a significant decrease of copia mRNA in lines with both the Gal4 driver and copia si/shRNA. The data underlying the graphs shown in the figure can be found in S1 Data , raw gel images can be found in S1 Raw Images . DLG = α-Discs Large (postsynaptic marker), HRP = α-horseradish peroxidase (presynaptic marker). N = from left to right; number of animals/NMJs quantified, 14/27, 9/17, 9/16 in (F) and (G). WT is C380-Gal4/Canton-S for C panels. ns p ≥ 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001. Abbreviations: Gag: Capsid forming proteins, which contains capsid (CA) and nucleocapsid (NC) sequences. Pol consists of the following subdomains: Protease (PR), Integrase (INT), reverse transcriptase (RT), and RNase H (RH). ENV: Envelope. The Long terminal repeat region (LTR) is needed for transposable element replication, consisting of U5, R, and U3 sub-regions. PBS: Primer Binding Site. PPT: Polypurine tract.

To study the function of Copia in Drosophila EVs, we first developed an antibody against the full-length, unspliced Copia protein (α-Copia Full ) that recognizes peptides encoded by the GAG and POL regions, and another antibody (α-Copia gag ) against a peptide encoded by Copia gag , an alternatively spliced isoform of the Copia transcript (Figs 1A and S1A ). Binding of α-Copia gag to Drosophila Schneider 2 (S2) cell lysates was blocked by the presence of the antigen peptide, validating the antibody specificity ( S1B Fig ). We probed lysates of the larval central nervous system (CNS) and body wall muscles (BWMs) and observed that the predominant band corresponds in size to the predicted molecular weight of Copia gag ( Fig 1B ). In contrast, in the BWM we see bands of varying sizes, which is consistent with auto-cleavage of the polyprotein. This is typically observed in retroviruses where the full-length polyprotein is cleaved into small functional peptides [ 13 – 16 ] and is consistent with the previously reported viral nature of TEs such as Copia [ 11 ]. These results indicate that the capsid-forming splice variant (Copia gag ) is abundant and the predominant form of Copia in larval CNS tissue.

Discussion

In this work, we present the first evidence that an active TE, Copia [30], is a potent regulator of structural synaptic plasticity. Interestingly, Copia and dArc1 seem to have an antagonistic relationship, providing evidence that the 2 capsid-encoding genes interact genetically to mediate plasticity at the fly NMJ. Copia is an inhibitor of synapse development, while dArc1 stimulates synapse formation. Proper synapse development therefore is a balance of these 2 opposing forces.

The details of this mechanism require further investigation. Copia controls plasticity, but is it dependent on the transfer of Copia capsids? We determined that EVs from S2 cells contain a detergent resistant electron dense structure that immunoreact with Copia antibodies, and further we found EVs immunoreacted with Copia antibodies, which strongly suggests that in S2 cells Copia is loaded into EVs presumably as a capsid. We also determined that presynaptic knockdown of Copia affects postsynaptic levels of Copia suggesting that Copia transfers across the synapse. However, work continues to determine if Copia, dArc1 or mammalian homologs of the latter transfer across synapses in a capsid in neurons. With the tools we have previously developed, and in this work, we will test if there are capsid-like structures at a synapse and whether they immunoreact with dArc1 and/or Copia antibodies, to confirm that the transfer of capsid-like structures is occurring at synapses.

The Copia knockdown phenotype is predominant to dArc1 for synaptic development at the larval NMJ, suggesting that dArc1 acts as an inhibitor of Copia, which is itself an inhibitor of plasticity. However, this does not reveal how this interaction might be carried out at the molecular level. A possible molecular mechanism is that each protein prevents assembly of the opposing capsid by binding to an assembly intermediate and blocking further oligomerization. We call this the “Poison Pill” hypothesis. We suggest this because both Copia and dArc1 form capsids using a related protein fold, which might indicate similar mechanisms. However, a challenge to this hypothesis is the distinct interaction surfaces within the Copia and dArc1 capsids (S5A Fig). Moreover, Copia and dArc1 do not co-immunoprecipitate, which indicates that they do not bind to each other or a shared intermediate simultaneously.

Another possible mechanism is that these capsids antagonize each other at the level of cargo transport. We believe that the currently available data is consistent with this model. For example, we show that Copia capsids require RNA for assembly, likely because Copia capsid proteins are weakly associated and the RNA acts as a “glue” to keep the capsid together. This would make Copia capsids exceptionally sensitive to the amount of RNA that is available. Furthermore, we observed that dArc1 and Copia have different RNA-binding capacities as determined by RIP-seq. Specifically, we see that dArc1 binds its own transcript but not Copia mRNA, while Copia binds its own transcript as well as dArc1 mRNA. Copia binding dArc1 mRNA this may sequester the dArc1 transcript from dArc1 capsids, thus limiting dArc1 function (S6I–S6K Fig). However, if dArc1 binds to its transcript with a higher affinity than Copia does, it would thereby protect dArc1 mRNA from Copia sequestration while reducing the available RNA substrate necessary for Copia capsid assembly (S6K Fig). Future experiments will investigate these possibilities.

Regardless of the nature by which Copia and dArc1 antagonize each other, the Copia structure suggests that it is substantially weaker than other retroviral or similar capsids. We observe that Copia capsids assemble through less extensive protein–proteins interactions than observed in most other retroelement-derived capsids. In general, capsid stability is controlled by the totality of interactions and the topological arrangement of the capsid subunits [31,32]. Because the topology of protein–protein interactions is identical across all retroelement-derived capsids, the total interaction area is likely to be a good proxy for overall capsid stability. Thus, we hypothesize that Copia capsids are less stable than most other retroelement derived capsids, thus requiring RNA to assemble. We note that the Ty3/Gypsy retrotransposon capsid has been documented to be relatively weak [25], indicating that weaker capsids might be a common feature of retrotransposons. It seems likely that the weaker capsids are related to the diverse morphologies observed in preparations of Copia and Gypsy capsids. It is possible that the weaker capsids are selected as part of the nuclear import mechanism. Klumpe and colleagues shows that the nuclear pore complex (NPC) selects for certain sizes of capsid to enter, thus we speculate that this weaker association may allow Copia capsids to re-arrange to better fit through nuclear pores [33].

Additionally, it is not clear how Copia is affecting plasticity and bouton formation. We observe postsynaptic Copia is dependent on presynaptic expression, also we observe that Copia forms capsids, and in S2 cells load these capsids into EVs. This suggests Copia forms a capsid, crosses the synapse, likely in an EV. Altogether, this leads to the exciting possibility that Copia may control the postsynaptic muscle in a non-cell autonomous manner by affecting processes in the nucleus such as transcription. A long-speculated way this could occur is by insertion of a TE into the genome. This could change the nearby chromatin structure or introduce novel transcriptional elements (i.e., promoters or repressors), thereby affecting gene transcription, which would be consistent with the somatic mosaicism described by Barbara McClintock [34]. This would require the full Copia machinery, including the reverse transcriptase and the integrase proteins. In support of this hypothesis, the Copia capsid structure shows pores like those of retroviral capsids that are used for transporting the dNTPs that are necessary for reverse transcription [20]. Our hypothesis is also supported by Klumpe and colleagues that resolves T = 9 Copia capsids in the nucleus [33]. It remains to be seen whether Copia inserts into the muscle genome, if these insertions are variant or not, and if this affects nearby gene expression. Alternatively, it may be possible for viruses or TEs [35] to affect gene expression without inserting into the genome through formation of nuclear aggregates. Regardless, the role of a TE in synaptic development needs to be further investigated to better understand synaptic plasticity as well as explore the potential functions of retrotransposons in directing host biology.

It is unlikely that Copia and dArc1 are the only capsid-encoding proteins that are transferred between cells, so there may be additional proteins that are functioning in a similar mechanism in the nervous system or other tissues. The Drosophila genome, like other eukaryotic genomes, has a huge repository of RTEs that could be used to modulate physiological functions with a capsid delivery-like mechanism similar to dArc1 and Copia. Recently, the mammalian capsid-like protein PEG10 was engineered to transfer RNA cargo, providing more evidence that the viral synaptic transfer of RNA (ViSyToR) pathway described in Ashley and colleagues is relevant beyond the Drosophila NMJ [36]. While the engineered capsid is direct evidence that PEG10 plays a role transporting RNA transcripts between cells, it suggests that a capsid-encoded gene can act in a viral-like manner. Together, these studies along with our discovery that the TE Copia regulates synaptic plasticity raises the likely possibility that dArc1 is one of a family of capsid-encoding genes that regulate cellular functions in neurons and throughout the body.

There are many examples of TE domestication, including dArc1, whereby a TE fragment is selected through evolution for its beneficial role for host fitness. However, until recently there have been few examples of an entire transposon is domesticated. Specifically, examples of this are the Cer1 gene of C. elegans and the Line-1 elements that have a role in early mouse embryonic development [37,38]. While it has long been speculated that TEs have a role in neuronal development, and there is data that expression of Line-1 in mammalian brains is developmentally regulated, no specific roles for Line-1 or any other TEs in mammalian neurons have been identified [39–41]. As such, the recent discovery of a TE regulating neuronal cell migration in zebrafish [42] and our discovery of a TE regulating synaptic formation and plasticity suggest that we are at the beginning of an exciting new era of research uncovering the physiological roles of enigmatic sequences such as TEs. While most previous studies of TE domestication have focused on late stages of domestication (such as with dArc1), our discovery of Copia’s role in synaptic plasticity opens the door to study domestication at earlier stages. What factors determine whether a full-length TE or a subset of a TE is domesticated? What does it mean to be domesticated? We envision that these and other exciting questions will begin to be addressed by the study of Copia.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002983

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/