Direct cell-to-cell transmission of retrotransposons

Abstract

Transposable elements are abundant in host genomes but are generally considered to be confined to the cell in which they are expressed, with the notable exception of endogenous retroviruses. Here, we identify a group of LTR retrotransposons that infect the germline from somatic cells within the Drosophila ovary, despite lacking the fusogenic Envelope protein typically required for retroviral entry. Instead, these elements encode a short transmembrane protein, sORF2, with structural features reminiscent of viral cell-cell fusogens. Through genetics, imaging, and electron microscopy, we show that sORF2 localizes to invasive somatic protrusions, enabling the direct transfer of retrotransposon capsids into the oocyte. Remarkably, sORF2-like proteins are widespread among insect retrotransposons and also occur in piscine nackednaviruses and avian picornaviruses. These findings reveal a noncanonical, Envelope-independent transmission mechanism shared by retrotransposons and non-enveloped viruses, offering important insights into host-pathogen evolution and soma-germline interactions.

Figure 1:. Somatically expressed 412 and Stalker2 RNA transmits into the oocyte.

(A) Phylogenetic tree of Metaviridae in D. melanogaster based on a multiple sequence alignment of the consensus pol region, rooted with Belpaoviridae as an outgroup. The Gypsy group (grey circle) includes both env-encoding insect ERVs and LTR retrotransposons lacking an env gene. The MDG1 group is highlighted in magenta. Scale bar: substitutions per site. (B) Expression fold change of transposable element (TE) groups measured by RNA-seq of ovaries upon somatic piRNA pathway knockdown (Tj>Gal4, vret-RNAi) compared to control (Tj>Gal4, arr2-RNAi) (n=3). The number of TEs in each group is indicated in brackets. (C) TE group distribution within TE-enriched heterochromatic tiles (100-kb genomic regions with >50% transposon content). Red datapoints represent flamenco tiles. Asterisks indicate significant deviation from the background distribution (MDG1 p-value = 0.0145). For (B-C) Gypsy-env refers to env-encoding insect ERVs, and Gypsy non-env refers to LTR retrotransposons without an env gene. Boxplots in (B-C) present the median (center line), IQR (box bounds), min and max within 1.5 IQR (whiskers), and outliers (points beyond whiskers). (D) Schematic of a Drosophila ovariole, showing sequential progression of follicles through the developmental stages of oogenesis in the adult ovary. Somatic follicle cells (green) are separate from the germline (beige), which is comprised of the oocyte and nurse cells. (E) Transcriptional LTR-lacZ reporters for 412 under different genetic conditions, (top) somatic piRNA pathway knockdown (Tj>Gal4, vret-RNAi); (middle) germline piRNA pathway knockdown (MTD>Gal4, aub-RNAi, ago3-RNAi); (bottom) LTR-lacZ reporter for germline-expressed Burdock TE under germline piRNA pathway knockdown (MTD>Gal4, aub-RNAi, ago3-RNAi). Scale bar: 100 μm. (F-H) Dual smFISH staining of early (F) and mid-oogenesis follicles (G-H) for LTR retrotransposons 412 (magenta) and Stalker2 (yellow) RNA. Panels (F) and (G, G’, G”) show somatic piRNA pathway knockdown (Tj>Gal4, vret-RNAi) while (H) shows the control knockdown for an unrelated gene (Tj>Gal4, arr2-RNAi). Panel (F) bottom shows zoom-in of boxed region in top image, at a different Z-section. White arrowheads denote RNA signals of LTR retrotransposons in the apical membrane of somatic cells in (F) and in the oocyte in (G’, G”). Somatic membranes are marked by myristoylated GFP. Scale bars (F-H): 10 μm.

Figure 2:. 412 and Stalker2 retrotransposons autonomously infect the oocyte.

(A) Schematic organization of the flamenco piRNA locus. The flam^Δ412-St2 allele was generated by removing a 15.5 kb region encompassing the 412-Stalker2 nested insertion. (B) Small RNA sequencing of piRNAs from flam^Δ412-St ovaries, compared to control ovaries (identical genetic background). Each datapoint represents the number of antisense reads mapped to a TE consensus sequence, normalized as reads per kilobase of TE per million mapped microRNAs (RPKM). (C) Volcano plot of whole transcriptome poly(A)-enriched RNA-seq, from flam^Δ412-St2 ovaries, compared to control ovaries (n=3). Each datapoint represents the number of reads mapped to a TE consensus sequence. (D-G) Dual smFISH of flam^Δ412-St2 ovaries for 412 RNA (magenta) and Stalker2 RNA (yellow) in (D) early developmental stages (1–7) (E) later developmental stages (7 and 9). (F-G) Zoomed-in regions from (E), showing absence of oocyte-localized signal for 412 and Stalker2 in stage 7, whereas in (G) stage 9 there is smFISH signal for both 412 and Stalker2 within the oocyte. (H) Whole-mount immunofluorescence of flam^Δ412-St ovaries, showing somatic apical accumulation and oocyte localization of 412 Gag proteins. (I) Same as (H), but for control ovaries (identical genetic background). For (D-I), GFP-labelled myosin II (green) is used to demarcate both the oocyte and soma membranes. Scale bar: (D-I) 10 μm.

Figure 3:. Characterization of the two unique sORFs, sORF1 and sORF2, encoded in 412 genome.

(A) Schematic of the genomic features encoded by MDG1 retrotransposons, exemplified by the 412 consensus sequence. Features include flanking LTRs, capsid-encoding gag and enzymatic machinery-encoding pol. Two short ORFs (sORF1 and sORF2), with a predicted premature transcription termination site (polyA) between them, are located upstream of gag. TSS, transcriptional start site. (B) Mapping of 3’ ends of long-read direct RNA reads corresponding to the 412 consensus sequence, in control ovaries (Tj>Gal4, arr2-RNAi, top) and somatic piRNA pathway knockdown ovaries (Tj>Gal4, vret-RNAi, bottom). The Y-axis indicates the number of reads, and the illustration represents the short and long isoforms produced based on the polyA site that is predominantly used. (C) Phylogenetic tree of the MDG1 clade based on pol alignment, with presence or absence of intact sORF1 and sORF2 sequences within the TE consensus sequences of D. melanogaster reference genome iso-1. Branch lengths are not to scale. (D) AlphaFold3 structural predictions for 412 sORF1 and sORF2, with cartoon representation color-coded by pLDDT confidence score (red: high, blue: low). (E) Notable features in sORF2 AlphaFold3 predicted structure: left, surface hydrophobicity (cyan: low, yellow: high); right, electrostatic surface potential (red: acidic, blue: basic). (F) Mass-spectrometry detection of peptides corresponding to sORF1 (purple), sORF2 (cyan) and Gag (orange) sequences in somatic piRNA pathway knockdown ovaries and OSCs. (G) Immunofluorescence staining of a stage 10 flam^Δ412-St2 follicle with α-sORF1 antibody (purple). (G’) Zoom-in of somatic nuclei in the boxed region in (G). (H) Co-localization of α-sORF2-N (cyan) and α-Gag (orange) on the apical somatic cell membranes in an early stage follicle. (I) Immunofluorescence signals of α-sORF2-C (cyan) and α-Gag (orange) often co-localize within the oocyte in later stage follicles. Right panels show zoomed-in region from I at a different Z-plane. For (G-I) F-actin is labelled with phalloidin (green) to demarcate cortical actin near the plasma membrane. Scale bars (G-I): 10 μm.

Figure 4:. sORF2 proteins resemble FAST proteins, and promote cell fusion by facilitating proximity of membranes.

(A) Multiple sequence alignment of sORF2 proteins from D. melanogaster and three previously described FAST proteins. p13 from bat Broome reovirus (BroV, ACU68609.1); p14 from reptilian orthoreovirus (RRV, AAP03134.1); p15 from baboon orthoreovirus (BRV, AAL01373.1). sORF2 proteins of Stalker1 and Stalker2 lack the predicted canonical N-myristoylation motif (N-myr, MGxxxS/T) but share all other features. Blood consensus sequence lacks an intact sORF2, but several heterochromatic insertions supported a Blood variant with the presented sORF2 sequence. Clustal X coloring scheme labels conserved residues according to amino acid profile, additional Arg and Lys residues in the polybasic region are shown in light red. (B) Left, Immunofluorescence Z-stack projection of stage 7 flam^Δ412-St2 follicles expressing GFP-labelled myosin II (green)stained with α-Gag (orange) showing accumulations of capsids within the somatic follicle cell layer covering the oocyte. Scale bar: 10 μm. Top right, zoom-in of the capsid accumulation pattern. Scale bar: 2 μm. Bottom right, mid-section of the follicle showing 412 capsid accumulations along the somatic apical membranes. Scale bar: 2 μm. (For C-D, F-G) Upper panels, Transmission (TEM) and immuno-EM images. Lower panels, schematic representation of EM images, somatic cells (green) with observed capsids (pink) are in close proximity to the oocyte membrane (beige) and sORF2 proteins (cyan). (C) TEM of capsid accumulations along somatic apical membranes in stage 7 flam^Δ412-St2 follicle. Scale bar: 500 nm. (D) Immuno-EM of flam^Δ412-St2 stage 7 follicle using α-sORF2-C antibodies, labelled with secondary antibody-conjugated gold particles (black dots in upper panel, cyan in lower schematic), revealing sORF2 localization near accumulated capsids. Scale bar: 100 nm. (E) Immunofluorescence of flam^Δ412-St2 ovaries co-stained with α-Gag (orange), α-sORF2-C (cyan) and phalloidin (green), showing co-localization in protrusions extending into the oocyte (white arrowheads). Scale bars: 10 μm, zoom-in panels on right: 2 μm. (F) TEM of an invasive protrusion filled with capsids in flam^Δ412-St2 follicle, extending into the oocyte. Adjacent physiologically normal spacing between membranes and typical microvilli are shown for comparison. Scale bar: 500 nm. Inset, zoom-in of protrusion with indistinguishable soma and oocyte membrane. (G) Immuno-EM of flam^Δ412-St2 ovaries using α-Gag antibody, with gold particles (black dots) corresponding to 412 capsids within an invading protrusion. Scale bar: 100 nm. (H) Snapshot of electron tomography segmentation and 3D rendering of an invasive protrusion in flam^Δ412-St2 ovaries, filled with 412 and Stalker2 viral-like particles (pink). The somatic membrane (green) is closely associated with the oocyte membrane (beige), with minimal separation between them. Scale bar: 100 nm. (I) Proposed model for soma-to-oocyte infectivity of MDG1 LTR retrotransposons. (Left) somatically expressed transmembrane sORF2 proteins (cyan) are closely associated with LTR retrotransposon capsids (pink). Upon local clustering and additional unknown cues, cortical actin-remodeling (green) is initiated (right), promoting the formation of invasive protrusions toward the oocyte membrane. Once the membranes are close enough in a fusogenic synapse structure, transient fusion between somatic and oocyte membranes enables LTR retrotransposons to enter the oocyte.

Figure 5:. sORF2/FAST-like sequences are abundant and widespread among non-enveloped viruses.

(A) Taxonomic tree of all viruses and TEs described in this study. (B) Taxonomic tree of arthropod genomes annotated in Repbase for Metaviridae consensus sequences. Inner circle (blue shades) – number of TE entries in the database per species. Outer circle (red shades) – number of sORF2/FAST-like consensus sequences found per species. Insect genomes analyzed in (D) are labelled with asterisk. (C) Schematic representation of computational pipeline to identify sORF2/FAST-like proteins in additional genomes. (D) Number of individual sORF2/FAST-like sequences in LTR retrotransposon insertions of three insect genomes. (E-F) Multiple sequence alignment of a subset of sORF2/FAST-like (E) smORF2 proteins found in fish-infecting nackednaviruses (F) ORF2 proteins from bird-infecting picornaviruses, and their typical genome organization. Clustal X coloring scheme labels conserved residues according to amino acid profile, additional Arg and Lys in the polybasic region are shown in light red.