Evolutionary diversification reveals distinct somatic versus germline cytoskeletal functions of the Arp2 branched actin nucleator protein

Summary

Branched actin networks are critical in many cellular processes, including cell motility and division. Arp2, a protein within the 7-membered Arp2/3 complex, is responsible for generating branched actin. Given its essential roles, Arp2 evolves under stringent sequence conservation throughout eukaryotic evolution. We unexpectedly discovered recurrent evolutionary diversification of Arp2 in Drosophila, yielding independently arising paralogs Arp2D in obscura species and Arp2D2 in montium species. Both paralogs are unusually testis-enriched in expression relative to Arp2. We investigated whether their sequence divergence from canonical Arp2 led to functional specialization by replacing Arp2 in D. melanogaster with either Arp2D or Arp2D2. Despite their divergence, we surprisingly found both complement Arp2’s essential function in somatic tissue, suggesting they have preserved the ability to polymerize branched actin even in a non-native species. However, we found that Arp2D- and Arp2D2-expressing males display defects throughout sperm development with Arp2D resulting in more pronounced deficiencies and subfertility, suggesting the Arp2 paralogs are cross-species incompatible in the testis. We focused on Arp2D and pinpointed two highly diverged structural regions—the D-loop and C-terminus—and found that they contribute to germline defects in D. melanogaster sperm development. However, while the Arp2D C-terminus is suboptimal in the D. melanogaster testis, it is essential for Arp2D somatic function. Testis cytology of the paralogs’ native species revealed striking differences in germline actin structures, indicating unique cytoskeletal requirements. Our findings suggest canonical Arp2 function differs between somatic versus germline contexts, and Arp2 paralogs may have recurrently evolved for species-specialized actin branching in the testis.

Graphical Abstract

In Brief

Stromberg et al. uncover testis-specific paralogs of the actin nucleator Arp2 in multiple Drosophila clades. The paralogs can perform Arp2 somatic roles in a non-native species, but sequence divergence leads to sperm development defects. Analysis of the native species suggests Arp2 diversification specialized for unique germline actin networks.

Introduction

The Arp2/3 complex is a highly conserved protein complex that nucleates branched actin networks and is found in almost all eukaryotes. The complex comprises seven proteins, consisting of two actin-related proteins, Arp2 and Arp3, and five other subunits, ArpC1–5. The complex docks onto a mother actin filament and promotes the nucleation of a daughter filament at a 70° angle from the mother filament¹. Arp2/3-generated branched actin networks are often found at the surface of cell membranes and are vital for many cellular processes, including motility, endocytosis^2–4, and signaling⁵. This complex also plays roles in genomic stability and cell division, particularly in proper chromosome segregation, spindle organization, and mitotic progression⁶. Moreover, Arp2/3 resides in the nucleus, where it aids in DNA damage repair^7,8.

In addition to its many somatic roles, Arp2/3 has specialized roles in the germline^9,10. Drosophila ovaries have Arp2/3-generated actin rings, which form channels between interconnected cells in the developing egg chamber and allow cytoplasmic flow to the oocyte⁹. Drosophila sperm development also depends on Arp2/3^11,12. Drosophila sperm develop together within cysts, sharing a cytoplasm, and mature sperm are separated following their elongation and tail development¹⁰. To accomplish ‘individualization,’ an actin cone forms around each sperm within the cyst, and all cones move synchronously along the tail toward the apical end of the testis^10,11. Arp2/3 localizes to the front of the cones, and branched actin polymerization aids in motility¹².

Recent studies have begun to reveal the structure and regulation of the Arp2/3 complex. Like actin, Arp2 and 3 contain four subdomains with an ATP-binding fold¹³. Upon docking onto the mother actin filament, Arp2/3 undergoes a conformational change from inactive to active, in which Arp2 and Arp3 are no longer splayed¹³ but aligned together, like actin monomers in a filament^14,15. This re-arrangement allows the Arps to become the first two subunits of the daughter actin filament^14,15. Arp2/3 cannot nucleate actin filaments on its own¹⁶; instead, nucleation-promoting factors (NPFs) activate Arp2/3 downstream of many signaling pathways. Numerous regulatory factors also terminate polymerization and disassemble filaments, replenishing the monomeric actin pool for future nucleation events¹⁷. Overall, many proteins regulate filament nucleation, polymerization, termination, and disassembly to fine-tune the timing and structural stability of branched actin networks¹⁷, thus augmenting the diversity of Arp2/3 functions.

Eukaryotic genomes show genetic and functional diversification of the Arp2/3 complex. Humans encode two isoforms of Arp3 (Arp3 and Arp3B), ArpC1 (ArpC1A and ArpC1B), and ArpC5 (ArpC5 and ArpC5L)^18,19. Arp2/3 complexes containing ArpC1B and ArpC5L differ in their rates of actin polymerization from those containing ArpC1A and ArpC5¹⁸. Although Arp3B functions similarly to canonical Arp3 in actin polymerization, it diverges from Arp3 in the rate of actin network disassembly¹⁹. Drosophila also exhibits Arp2/3 diversification. D. melanogaster encodes two isoforms of Arpc3 (Arpc3a and Arpc3b). Arpc3a is expressed in all tissues, whereas Arpc3b appears to be expressed only in the ovary and localizes to actin rings⁹. Furthermore, an Arp2 paralog, known as Arp2D, emerged in one clade of Drosophila species. Arp2D is solely expressed in the male germline and localizes to branched actin networks in cones²⁰. Key regulators of Arp2/3 also show diversification, allowing for differential regulation between cell types^17,21. Therefore, despite Arp2/3’s stringent conservation, multiple members of the Arp2/3 complex and its regulators have undergone recurrent gene duplication and diversification across phyla, suggesting that evolution is selecting for different ‘flavors’ of Arp2/3 for specialized roles.

Here, we show that Arp2 has undergone recurrent diversification in Drosophila. In addition to Arp2D in the Drosophila obscura clade²⁰, we discovered a second, independently arising gene duplication of Arp2 in the montium clade: ‘Arp2D2.’ Both Arp2D and Arp2D2 are exclusively expressed in the testis of their respective species. Despite their sequence divergence, we found that the paralogs complement essential somatic functions of canonical Arp2 in the non-native species D. melanogaster. However, the paralogs fail to fully complement Arp2 function in the testis with prominent defects in actin cones, suggesting they are incompatible with D. melanogaster sperm development. We investigated the sequence divergence of Arp2D, which localizes to actin cones in a native species²⁰, and found that the divergent D-loop and C-terminus of Arp2D contribute to cone defects in D. melanogaster. However, the unique C-terminus is critical for Arp2D somatic function. Interestingly, actin cones in the native species that express Arp2D and Arp2D2 are structurally distinct from those in D. melanogaster. Overall, our data suggest that Arp2 paralogs have evolved for species-specialized testis actin branching, and their divergence has led to cross-species incompatibility in the germline.

Results

Recurrent invention of testis-expressed Arp2 paralogs in Drosophila species

The seven-member Arp2/3 complex evolves under stringent sequence constraints in most eukaryotes. However, the Arp2 subunit has undergone unexpected diversification via gene duplication. We previously discovered that an Arp2 paralog, Arp2D, arose via retroduplication in the obscura clade and has been retained for over 14 million years²⁰. By examining additional sequenced Drosophila genomes, we found that D. kikkawai²² and D. serrata²³, species of the montium clade, also encoded an Arp2 gene duplication found on chromosome 2 (Figure 1A), and we named it ‘Arp2D2.’ We queried the Arp2D2 locus by PCR in 18 additional montium species whose genomes were unsequenced. Together with recently published montium genomes²⁴, we concluded that Arp2D2 is present in a shared syntenic location in all 27 surveyed montium species (Figure 1B and S1A). Based on this, we conclude that Arp2D2 arose in the common ancestor of the montium group at least 15 million years ago²⁵ (Figure S1A). Our phylogenetic analysis based on a nucleotide alignment shows that Arp2D2 sequences form a monophyletic clade that is a well-supported sister clade to the montium Arp2 orthologs, further confirming Arp2D2 duplicated from canonical Arp2 in the common ancestor of the montium group (Figure 1B). Arp2D orthologs form a separate clade with its closest outgroup being obscura Arp2 sequences, supporting independent origins of the paralogs (Figure 1B). Arp2 encodes six introns (Figure 1A), yet 26 of the 27 Arp2D2 sequences have none (D. diplacantha Arp2D2 subsequently acquired one intron (Figure S1B and Data S1A)). Therefore, our phylogenomic analyses reveal that Arp2D2 and Arp2D both arose via independent retroduplication events in which Arp2 spliced mRNA was reverse transcribed and inserted into the genome²⁰ (Figure 1A and 1B). Following their origin, both Arp2D and Arp2D2 have been retained throughout the obscura and montium clades, respectively, suggesting that they confer a functional advantage to their native species.

Figure 1: Arp2D2 is a male germline-enriched Arp2 paralog found in the montium group species of Drosophila.

(A) Syntenic loci of canonical Arp2 in representative Drosophila species spanning 50 million years of evolution. Arp2D, found in the obscura clade, was described previously²⁰. Arp2D2 was discovered in montium group species. (B) Phylogenetic tree derived from Arp2, Arp2D, and Arp2D2 coding sequences with 100X resampling. Blue asterisks indicate Arp2D2 sequences obtained from sequenced genomes^22–24. Remaining Arp2D2 sequences were obtained from this study (see also Data S1A). Arp2D and Arp2 sequences were obtained from a previous study²⁰ and sequenced genomes^22,24,44, respectively. Ranges of protein identity among the orthologs and nodes with >80% bootstrap support are noted. Stars indicate independent evolutionary origins of Arp2D2 and Arp2D. (C) RT-PCRs comparing mRNA expression of Arp2D2 between females and males in four montium species. Rp49 confirmed similar cDNA quantities among the samples. (D) RT-PCRs comparing Arp2D2 mRNA expression in D. auraria whole females (WF) and males (WM). Males were further dissected to localize Arp2D2: head (H), remaining carcass (C), and testes (T). Arp2 indicates presence of cDNA among the samples. See also Figure S1 and Table S1–3.

Although independently derived from canonical Arp2, Arp2D and Arp2D2 may have converged on similar sequence features associated with their testis-specific roles. However, we observed no conserved amino acid changes common to all Arp2D and Arp2D2 orthologs but distinct from canonical Arp2, suggesting the paralogs had different evolutionary paths (Figure S1C). Moreover, when we viewed the conserved residue changes on the surface of the Arp2 structure (where protein interactions occur), we did not find a ‘hot spot’ where Arp2D or Arp2D2 residue changes concentrated (Figure S1C). Although Arp2D is rapidly evolving²⁰, we found that Arp2D2 is not subject to positive selection (Figure S1D and S1E), further indicating distinct evolutionary paths.

Arp2D exhibits testis-enriched expression, unlike the ubiquitously expressed canonical Arp2²⁰. We investigated whether Arp2D2 also exhibits sex-biased expression. We compared Arp2D2 mRNA expression between males and females in the montium species D. barbarae, D. bicornuta, D. watanabei, and D. auraria by reverse-transcription PCR analyses. We detected Arp2D2 only in males (Figure 1C, S1F and S1G). To determine if the expression is testis-specific, we extracted RNA from dissected D. auraria tissue and detected Arp2D2 mRNA solely in the testis, not the head or remaining carcass (Figure 1D, S1H, and S1I). Thus, two independent gene duplications of Arp2 exhibit testis-enriched expression, suggesting that Arp2 paralogs have recurrently acquired roles in sperm development.

Divergent Arp2 paralogs localize to actin in D. melanogaster

Our previous study revealed that GFP-tagged Arp2D, under the control of its endogenous promoter, localizes to actin, including actin cones, in the testis of its native species, D. pseudoobscura²⁰. Cones are testis-specific structures specialized for Drosophila sperm development and facilitate the separation of interconnected mature sperm¹⁰ (Figure 2A). It was unclear whether a divergent Arp2 paralog like Arp2D or Arp2D2 could localize to actin in a heterologous species. To investigate this possibility, we tested whether D. pseudoobscura Arp2D (from the obscura group) or D. auraria Arp2D2 (from the montium group) could localize to actin cones in D. melanogaster (D. mel). We generated D. mel fly lines that express the paralogs tagged with superfolder GFP²⁶ (sfGFP, Figure 2B). We expected the paralogs to be expressed in both somatic and germ cells because they are under the control of the endogenous Arp2 promoter (Figure 2B). Indeed, Arp2D, Arp2D2, and canonical Arp2 from D. pseudoobscura were expressed in both cell types and localized to actin cones, even in the presence of endogenous D. mel Arp2 (Figure 2C and S2A). We also observed that Arp2D and Arp2D2 distribute along the cone’s length with signal highest near the front like Arp2 (Figure 2C and S2B). Based on their localization, we conclude that Arp2D and Arp2D2 fold and localize to actin, suggesting they incorporate into the D. mel Arp2/3 complex despite sequence divergence.

Figure 2: Divergent Arp2 paralogs localize to actin in D. melanogaster.

(A) Schematic depicting individualization of mature sperm with actin cones. Each sperm within a cyst has one cone. All cones move synchronously from the nucleus to the end of the tail, encompassing each sperm with its own membrane and moving out excess cytoplasm (‘cystic bulge’). Cones contain parallel actin filaments in the rear and Arp2/3-generated branched actin networks in the front half. (B) Schematic showing genetic modification of wildtype D. mel flies, which encode Arp2 on the X-chromosome. Canonical Arp2 (D. pseudoobscura), Arp2D (D. pseudoobscura), or Arp2D2 (D. auraria) were C-terminally tagged with superfolder GFP²⁶ (sfGFP) and inserted on the third chromosome. All transgenes were under the control of the endogenous Arp2 promoter. (C) Testes from lines in (B) were fixed and probed for GFP and actin (about 20 testes imaged per genotype). Representative slices of cones are shown, and the merge of actin (magenta) and GFP (green) appears white if levels are similar. See also Figure S2 and Data S1.

Divergent Arp2 paralogs rescue somatic roles

Mutations or knockdowns of Arp2/3 complex components lead to lethality due to essential roles in somatic tissue^3,9. We tested if the Arp2D and Arp2D2 paralogs could rescue loss of D. mel Arp2. We first generated an Arp2-knockout (KO) fly line by deleting the entire gene and replacing it with an eye-expressed DsRed cassette using CRISPR/Cas9 technology (Figure 3A). Upstream to DsRed in the KO locus, we introduced an attP site, which we subsequently used for targeted transgenesis. As expected, our Arp2-KOs exhibited homozygous lethality. Because Arp2 is on the X chromosome, only heterozygous Arp2-KO females survived, whereas hemizygous Arp2-KO males did not.

Figure 3: Arp2 paralogs rescue Arp2 somatic functions in D. melanogaster, yet Arp2D-expressing males are subfertile.

(A) Schematic of CRISPR strategy for generating Arp2 knockouts (KOs). Cut sites were in the intergenic regions upstream and downstream of D. mel Arp2. DsRed, under the control of an eye-specific promoter, was inserted to track the KO allele. An attP site upstream of DsRed was leveraged for site-directed transgenesis. Tagless and intronless D. melanogaster Arp2, D. pseudoobscura Arp2D, or D. auraria Arp2D2 were inserted at the attP site, and sfGFP was expressed in the eye as a transgene marker. (B) Left, schematic of crossing scheme for testing rescue of Arp2-KO lethality. Heterozygous females (transgene balanced with FM7) were crossed to hemizygous males. Eye color indicates the transgene, with heterozygous females having dimmer GFP-positive eyes. Right, percent of the progeny population without FM7 (homozygous females and hemizygous males) shown for Arp2, Arp2D, and Arp2D2 (6–8 replicates each). Genotype fractions were compared to Mendelian expectation using a chi-squared test. Homozygotes were higher than expected, suggesting a slight fitness cost from FM7. (C) Male fertility assay conducted by crossing males from (A) to wildtype D. mel females. Progeny count differences were not significant. (D) Male fertility assay as in (C) except under heat stress (29°C). Arp2D-expressing males produced significantly less progeny than Arp2- or Arp2D2-expressing males. (E) Left, seminal vesicles (outlined in yellow) from Arp2-and Arp2D-expressing males from (A) imaged with a DNA probe. Right, area was measured, revealing significantly smaller seminal vesicles from Arp2D-expressing males (non-virgin). See also Figures S2, S3 and Data S1.

We next inserted D. pseudoobscura Arp2D, D. auraria Arp2D2, or canonical D. mel Arp2 (a positive control) in the Arp2-KO locus to test for rescue. Because Arp2D and Arp2D2 lack introns, we used intronless D. mel Arp2 to prevent phenotypes arising from splicing differences. All transgenes were introduced using the same attP site, and they all possess the same upstream and downstream ~1kb regions of D. melanogaster canonical Arp2, which included the 5’- and 3’ UTRs (Figure 3A). Because Arp2D is only 70% identical to Arp2, we expected Arp2D would fail to rescue lethality, whereas Arp2D2, being 96% identical, was likely to rescue. We were surprised to obtain homozygous females and hemizygous males expressing Arp2D or Arp2D2, suggesting both rescued Arp2-KO lethality. However, whether the transgene fully rescued lethality remained unclear. To address this question, we crossed females and males that encode one copy of the transgene and quantified the progeny genotypes (Figure 3B). This scenario should lead to 50% of females having two copies of the transgene and 50% of males having one copy if there is no fitness cost. Indeed, we found that crossing both Arp2D- and Arp2D2-expressing flies led to at least 50% of the progeny being homozygous (female) or hemizygous (male) for the transgenes (p<0.05, Figure 3B). The slight increase over the expected 50% suggests that the X-chromosome balancer (FM7) has a small fitness cost compared to the transgenes. We conclude that when canonical Arp2 is replaced with either Arp2D or Arp2D2, no loss of fitness is detected under laboratory conditions (Figure 3B). We further tested whether a fitness cost would become apparent under environmental stress. We conducted the same cross with the most diverged paralog Arp2D under heat stress at 29°C and found complete rescue, suggesting no fitness cost in viability even under these conditions (Figure S2C). We also did not notice lifetime defects, morphological abnormalities, or unusual behavior at room temperature or 29°C, though subtle defects in somatic tissues may exist. We conclude that the diverged Arp2D and Arp2D2 paralogs can functionally replace Arp2 for all assayed somatic roles, even in a non-native species.

Arp2D-expressing males are subfertile

Having established that Arp2D and Arp2D2 possess Arp2 somatic functions in D. mel, we next tested if they could fully complement Arp2 male germline functions by comparing fertility of the ‘gene-replacement’ males. We crossed hemizygous males with wildtype females and, after one week of mating, compared adult progeny count to males encoding canonical Arp2 in the KO locus (Figure 3C). We found that Arp2D- and Arp2D2-expressing flies yielded progeny counts comparable to Arp2-expressing flies (Figure 3C). Since males generally produce many sperm, one week of mating at room temperature may not reveal decreases in sperm count. Heat stress often exacerbates existing fertility defects and renders them detectable. We conducted the same male fertility assay at 29°C, and we found that Arp2D-expressing males exhibited a significant decrease in adult progeny (p=0.02, Figure 3D), indicating they are subfertile while Arp2D2-expressing males did not show a significant difference in progeny count from Arp2-expressing flies. We then directly assessed sperm production by measuring the size of the seminal vesicle, where mature sperm are stored²⁷. The area of the seminal vesicle directly correlates with the number of stored mature sperm²⁷. We found that the average area of seminal vesicles from Arp2D-expressing males at room temperature was significantly smaller than those from Arp2-expressing males, whether they mated or not (p=0.009, Figure 3E and S2D). The average area of Arp2D2-expressing seminal vesicles showed no reduction in size (Figure S2D). Therefore, Arp2D-expressing males exhibit a fertility cost and produce less sperm even under unstressed conditions. We investigated whether the decrease in germline function was due to a lack of Arp2D mRNA in the testis and found that Arp2D and Arp2D2 are both expressed in the testis at comparable levels (Figure S2E). Given Arp2D’s mRNA expression and the localization of Arp2D-GFP to testis actin (Figure 2C), we determined that expression levels cannot account for the observed male subfertility. We conclude that Arp2D may be able to perform somatic roles in D. mel yet lacks compatibility with D. mel sperm development.

Given the fertility cost in Arp2D-expressing males, we also explored the fertility of Arp2D-expressing females. We crossed homozygous females to wildtype males and found that Arp2D-expressing females exhibited a significant decrease in adult progeny (p<0.001, Figure S3A). Cytological analysis revealed an Arp2/3 loss-of-function phenotype⁹, in which ovarian actin structures (‘ring canals’) were misshapen (Figure S3B). We detected Arp2D mRNA in the ovary, suggesting this phenotype is not due to a lack of Arp2D expression (Figure S3C). We imaged ovaries from flies encoding Arp2D-sfGFP (Figure 2B) and found Arp2D-sfGFP, although detectable by immunoblot, did not appear to localize to actin, in contrast to D. pseudoobscura Arp2-sfGFP and D. auraria Arp2D2-sfGFP (Figure S3D and S3E). Together, these data suggest that Arp2D fails to function in ovaries, further supporting that despite its ability to perform somatic roles, Arp2D is incompatible with the D. mel germline. Because Arp2D is natively expressed exclusively in the testis²⁰ and notably localizes to testis actin even in a non-native species (Figure 2C), we focused on investigating protein function of the Arp2 paralogs in the male germline.

Arp2D- and Arp2D2-expressing males exhibit testis defects

We next investigated whether the testis of Arp2D-expressing flies exhibits cytological defects that may lead to reduced sperm production. Sperm development can be tracked spatially in the testis (Figure 4A). Germline stem cells are at the apical end, where they differentiate and progress through mitosis and meiosis. After sperm elongation, actin cones form at the basal end and progress toward the apical end, separating sperm within the same cyst (‘individualization,’ Figure 4A). Fly testes normally appear as tubes coiled at the basal end, yet Arp2D-expressing testes exhibited a striking morphological defect in which the apical end appeared swollen relative to the rest of the testis (p<0.0001, Figure 4B). Because Arp2D was not codon-optimized for D. mel, we considered this cytological defect might result from transcriptional and translational differences. However, codon-optimized Arp2D-expressing males exhibited the same abnormal morphology (Figure S4A and S4B). We also ruled out the possibility of an overexpression artifact, as testes expressing multiple copies of both Arp2D and endogenous Arp2 display wildtype morphology (Figure S4C).

Figure 4: Arp2D- and Arp2D2-expressing D. melanogaster males exhibit defects in late spermatogenesis.

(A) Schematic showing the stages of sperm development progressing from the apical to the basal end. Nuclei are depicted in blue. (B) Left, brightfield images of Arp2- and Arp2D-expressing testes. Right, Arp2D-expressing flies exhibit an enlarged apical end, quantified as percent abnormal (Arp2: n=19; Arp2D: n=12; Arp2D2: n=23 testes). (C) Enlarged images of actin cones from fixed Arp2- or Arp2D-expressing testes, stained for actin. Cones are fan-like due to branched actin (see cartoon). Images are maximum projections showing the entirety of a few cones; because cones are hollow, they occasionally appear as doublets. Cones that were fully formed, synchronous, and mid-individualization were measured at the leading edge, indicated by brackets. Arp2D-genereated cones were significantly narrower (Arp2: n=35 cones across 8 testes; Arp2D: n=49 cones across 9 testes). (D) Left, images of all cones in one cyst. Arp2-expressing testes exhibit synchronous cones while Arp2D-expressing testes have smaller, trailing cones (see asterisk and corresponding inset). Right, quantification of cysts with asynchronous cones (Arp2: n=33 cysts across 11 testes, Arp2D: n=29 cysts across 9 testes). (E) Image of the apical end of Arp2- or Arp2D-expressing testes, stained for actin. In Arp2D-expressing testes, elongated cysts are disorganized, curving back toward the basal end (see inset). (F) Left, image of the apical end in fixed Arp2D-expressing testis, stained for actin. Motile cones abnormally face the basal end. Arrow indicates direction cone front is facing. Right, quantification of cysts with cone fronts facing the basal end (Arp2: n=11; Arp2D: n=9 testes). (G) Quantification of the number of waste bags per testis. Arp2D-expressing testes exhibited significantly more waste bags (Arp2: n= 11; Arp2D: n=9 testes). (H) Left, images of fixed and stained Arp2- or Arp2D2-generated actin cones. Brackets indicate leading edge measured for width. Right, cone widths were quantified as in (C), indicating Arp2D2-generated cones are narrower (Arp2: n=63 cones across 21 testes; Arp2D2: n= 78 cones across 25 testes). (I) Left, image of cones from Arp2D2-expressing testes. Asterisk indicates asynchronous cones. Right, quantification of the number of cysts with asynchronous cones (Arp2: n=55 cysts across 17 testes; Arp2D2: n=108 cysts across 38 testes). (J) Quantification of the number of waste bags per testis, indicating a significant increase with Arp2D2 expression (Arp2: n =17, Arp2D2: n =38 testes). See also Figure S4, S6 and Data S1.

The enlarged apical end of Arp2D-expressing testes is reminiscent of phenotypes previously observed in some sperm individualization mutants, which exhibited scattered actin cones and an accumulation of debris and disorganized cysts at the apical end^28–32. Because of these mutants^28–32 and Arp2D localization to cones in a native species²⁰, we investigated Arp2D-expressing testes for defects in D. mel individualization. We fixed and probed testes for DNA and actin to visualize cones in near-mature sperm. Arp2D-generated actin cones exhibited the typical fan shape that indicates the presence of the branched actin network in the front half¹² (Figure 4C). However, we measured the leading edge of fully formed, motile cones (no longer at sperm heads) and found the front half was narrower than Arp2-generated cones (p<0.0001, Figure 4C). We also observed smaller, trailing cones, indicating cones were moving asynchronously (p<0.0001, Figure 4D), unlike wildtype cones¹². Cysts with mature sperm were also disorganized; the end of a cyst is generally found at the apical end, but we observed mature cysts in the bulbous tip curving back toward the basal end (Figure 4E and S4D). Due to the disorganization of cysts, cones were frequently misoriented with cone fronts facing the basal end instead of the apical end (p=0.05, Figure 4F). We next assessed the last step of individualization; once actin cones reach the apical end, the excess cytoplasm and cones are degraded and become a ‘waste bag’¹⁰. We found that the number of waste bags per testis significantly increased (p=0.005, Figure 4G) and was not due to a lack of caspase proteolytic activity (Figure S4D), a requirement for waste bag removal and proper individualization^30,33,34. Overall, Arp2D-expressing males exhibited prominent defects throughout late spermatogenesis.

Although Arp2D2-expressing D. mel males appeared fertile (Figure 3C and 3D), we investigated whether the testis had defects similar to those produced by Arp2D. We found that Arp2D2-expressing testes displayed wildtype morphology (Figure 4B) and did not exhibit disorganized cysts or misoriented actin cones (Figure S6B and S6D). However, cones were significantly narrower than Arp2-generated cones (p<0.0001, Figure 4H), and cysts often displayed asynchronous cones (p<0.0001, Figure 4I). We also found a significant increase in waste bags per testis (p=0.05, Figure 4J). Therefore, Arp2D2 leads to defects in individualization, yet these phenotypes, which are less pronounced than the effects in Arp2D-expressing testes, do not reduce male fertility or sperm production (Figure 3C, 3D, and S2D). Because Arp2D2-expressing testes are morphologically wildtype yet display cone defects, the abnormal morphology of Arp2D-expressing testes is likely due to the accumulation of disorganized cysts, not asynchronous cones, suggesting the many germline defects resulting from Arp2D are pleiotropic. Overall, these data suggest that Arp2D and Arp2D2 fail to carry out sperm individualization optimally in D. mel, and the more diverged Arp2D is most incompatible with D. mel sperm development. Therefore, we infer that the Arp2 paralogs have likely specialized for sperm development in their native species.

The divergent D-loop contributes to testis defects

We next focused on Arp2D and asked which sequence divergence leads to testis defects in D. mel. We compared the protein sequences of Arp2D orthologs to canonical Arp2 and Arp2D2 orthologs and found the most divergence in two segments: the D-loop and C-terminus (Figure S5A–C). Canonical Arp2 has a highly conserved alternatively spliced exon in the D-loop, a flexible region in subdomain 2 (Figure S5B and S5C). The D-loop in Arp2D2 orthologs appears similar in length and sequence to the longer Arp2 splice variant, whereas Arp2D orthologs do not encode the alternative exon and diverge in sequence (Figure S5A and S5B). Most Arp2D2 orthologs have C-termini similar to canonical Arp2, yet Arp2D orthologs have longer and diverged C-termini ranging from 10 to 21 additional charged amino acids, composed mainly of Lys and Asp/Glu (Figure S5B). Given canonical Arp2’s conservation across eukaryotes, this sequence divergence may result in functional divergence.

We first investigated if the unique D-loop contributes to defects in sperm development. We tested if we could rescue some of the testis phenotypes, including abnormal morphology, produced by Arp2D by replacing its D-loop with the D-loop from canonical Arp2 or Arp2D2, which exhibited fewer defects. We were unable to obtain a chimera with the Arp2 D-loop but successfully generated Arp2D-expressing flies with the D-loop from D. auraria Arp2D2 (Figure 5A and S5D). Similar to previous designs, the chimera encoded the canonical Arp2 5’ and 3’UTRs and was inserted into the Arp2-KO locus to assess its impact on testis function (Figure 3A). The chimera’s expression was comparable to Arp2 (Figure S6H) and fully rescued Arp2-KO lethality (Figure 5B). Moreover, while 100% of Arp2D-expressing testes were bulbous, only 30% of the chimera’s testes exhibited a bulbous apical end (p<0.001, Figure 5C and S6B). However, even more cysts exhibited asynchronous cones than in Arp2D-expressing testes (p<0.001, Figure 5D and S6C), and fully formed, non-trailing cones were as narrow as Arp2D-generated cones (Figure 5E and S6E). The number of waste bags per testis and orientation of cones was comparable to Arp2D-expressing males (Figure S6A and S6D). We conclude that the diverged D-loop in Arp2D contributes to the pleiotropic effects in the germline, particularly disorganized cysts, yet it is not the only sequence divergence leading to defects. Furthermore, given the chimera resulted in even more cysts with asynchronous cones, the Arp2D2 D-loop likely leads to asynchronous cones observed in Arp2D2-expressing testes.

Figure 5: The divergent D-loop contributes to testis defects.

(A) Schematic showing how the chimera ‘Arp2D-2D2 loop’ was generated: the D. pseudoobscura Arp2D D-loop was swapped with the D. auraria Arp2D2 D-loop. Chimeric Arp2D was inserted into the D. mel Arp2-KO locus as in Figure 3A. (B) Schematic of the cross testing for rescue of Arp2-KO lethality with ‘Arp2D-2D2 loop’ as in Figure 3B. Progeny genotype fractions were compared to Mendelian expectation using a chi-squared test. The chimera fully rescued Arp2-KO lethality (5 replicates). (C) Left, brightfield images of one-day old virgins expressing Arp2D or ‘Arp2D-2D2 loop.’ Arrowheads indicate apical ends. Right, quantification of the percent of testes with an abnormal (bulbous) apical end (chimera: n=28 testes; Arp2 and Arp2D data from Figure 4B). Most chimera-expressing testes exhibited normal apical ends. (D) Quantification of the percent of cysts containing asynchronous cones (chimera: n=15 cysts across 8 testes; data for Arp2 and Arp2D from Figure 4D). The chimera exhibits the most asynchronous cones. (E) Cones were measured as in Figure 4C. Chimera-generated cones were comparable in width to Arp2D-generated cones (Arp2D-2D2 loop: n=52 across 9 testes, Arp2 and Arp2D: same as Figure 4C). (F) Schematic showing how the chimera ‘Arp2-2D loop’ was engineered: the Arp2 D-loop was swapped with the Arp2D D-loop. ‘Arp2-2D loop’ was inserted into the D. mel Arp2-KO locus. (G) Schematic of the cross used to test for rescue of Arp2-KO lethality with ‘Arp2-2Dloop.’ Progeny genotype fractions were quantified as in Figure 3B, and the chimera fully rescued Arp2-KO lethality (5 replicates). (H) Quantification of the percent of cysts containing asynchronous cones, indicating significantly more in chimera-expressing testes. Chimera: n=60 cysts across 14 testes, Arp2: n=53 cysts across 13 testes. (I) Cone widths were quantified as in (E). Chimera-generated cones are narrower than Arp2-generated cones (Arp2: n=60 cones across 13 testes, Arp2-2D loop: n=70 cones across 19 testes). See also Figure S5, S6 and Data S1.

To directly probe how divergence of the Arp2 D-loop impacts germline function, we replaced the D-loop in canonical Arp2 with the diverged Arp2D loop. Using the same approach, we generated the chimera ‘Arp2-2D loop’ (Figure 5F and S5E). Although the chimera rescued Arp2-KO lethality and led to normal testis morphology (Figure 5G, S6B, and S6G), we detected an increase in cysts with asynchronous cones, and cones were narrower than Arp2-generated cones (p<0.001, Figure 5H and 5I). These data further support the conclusion that D-loop divergence leads to testis incompatibility in a non-native species.

The Arp2D C-terminus is required for activity in somatic tissue but reduces optimal testis function

We next tested if Arp2D’s unique C-terminus also affects sperm development. We shortened the D. pseudoobscura Arp2D C-terminus by removing 11 amino acids, including the lysines conserved among most obscura species, and replaced it with the canonical Arp2 C-terminus, which is three residues long (Figure 6A and S5F). We inserted the chimeric transgene (‘Arp2D-2CT’) into the Arp2-KO locus similar to the previous chimeras, and although we removed divergence in the Arp2D C-terminus, Arp2D-2CT surprisingly failed to rescue Arp2-KO lethality (p<0.0001, Figure 6B). We verified that the construct was expressed and comparable to fulllength Arp2D expression in heterozygotes (Figure S6F). Therefore, the evolutionarily unique C-terminus in Arp2D is critical for Arp2D activity.

Figure 6: The unique C-terminus is required for Arp2D function but leads to testis defects in D. melanogaster.

(A) Schematic showing the generation of two chimeras. The Arp2D C-terminus was replaced with Arp2’s short C-terminus, producing ‘Arp2D-2CT.’ The Arp2D C-terminus was engineered on to Arp2, forming the ‘Arp2-2DCT’ chimera. Transgenes were inserted into the Arp2-KO locus as in Figure 3A. (B) Schematic of the cross testing for rescue of Arp2-KO lethality with Arp2D-2CT. Percent of male progeny that was hemizygous for Arp2D-2CT is shown. No viable hemizygous males were produced (5 replicates). (C) Schematic of the cross testing for rescue of Arp2-KO lethality with Arp2-2DCT. Progeny genotype fractions in B and C were compared to Mendelian expectation using a chi-squared test. Arp2-2DCT fully rescued Arp2-KO lethality (6 replicates). (D) Percent of abnormal testes (bulbous apical end). Arp2-2DCT-expressing flies exhibit an increase in abnormal testes (Arp2-2DCT: n=31 testes; data for Arp2 from Figure 4B). (E) Percent of cysts containing asynchronous cones. Arp2-2DCT-expressing testes exhibit significantly more asynchronous cones (Arp2-2DCT: n=39 cysts across 15 testes; Arp2 same as Figure 4D). (F) Leading edges of non-trailing cones were measured. Arp2-2DCT-generated cones were significantly narrower than Arp2-generated cones (Arp2-2DCT: n=49 cones across 11 testes; Arp2: same as Figure 4C). See also Figure S5, S6 and Data S1.

We next tested the impact of the Arp2D C-terminus on canonical Arp2 function. We modified the Arp2-KO flies by inserting Arp2 with Arp2D’s C-terminus (‘Arp2-2DCT,’ Figure 6A and S5G). The transgene was expressed (Figure S6H), and Arp2-KO lethality was fully rescued (Figure 6C). However, we found a significant increase in testes with an enlarged apical end (p=0.01, Figure 6D) and more cysts with asynchronous actin cones (p=0.01, Figure 6E). Cones were also narrower than Arp2-generated cones (p<0.0001, Figure 6F). Therefore, despite the importance of the unique C-terminus for Arp2D function, this domain negatively impacts D. mel sperm development and suggests a role in species-specific function.

Obscura and Montium species’ actin cones have diverged from D. melanogaster cones

Given heterologous expression of Arp2D and Arp2D2 in a non-native species leads to actin cone defects, we investigated whether cones in the paralogs’ native species exhibit distinct features. We compared actin cones across the species D. pseudoobscura (D. pse), D. auraria (D. aur), and D. mel and found that D. pse, the most evolutionarily distant, exhibits cones with the most diverged architecture (Figure 7A). D. pse cones were longer with a distinct front half (Figure 7A and 7B), the region composed of branched actin networks (Figure 2A). Unlike D. mel and D. aur cones, actin appears to project beyond the leading edge of D. pse cones (Figure 7B). D. aur exhibits cones most similar to D. mel cones (Figure 7A and 7B); however, D. aur and D. pse cones are narrower than D. mel cones (p < 0.0001, Figure 7C). Interestingly, D. aur and D. pse cone widths are similar to those in Arp2D- and Arp2D2-expressing D. mel flies (Figure 4C and 4H). Because Arp2D localizes to the front half of D. pse actin cones²⁰, our findings suggest Arp2D generates narrower branched actin networks within cones, even in a non-native species (Figure 4C). Although Arp2D2 localization in the D. aur testis is unknown, we hypothesize, based on its native testis expression and the resulting cone structure in D. mel, that Arp2D2 likely contributes to the generation of narrower cones in its native species. Overall, the divergence of obscura and montium cones suggests these species have unique cytoskeletal requirements for individualization that are absent in D. melanogaster.

Figure 7: Actin cones differ in structure across Drosophila species.

(A) Actin cones stained with the probe silicon-rhodamine (SiR)-actin⁴⁵ in the species D. melanogaster, D. auraria, and D. pseudoobscura. Relative evolutionary distances are indicated with the above phylogenetic tree, and the corresponding paralog encoded by D. aur and D.pse is noted. (B) Closer views of cones in (A). Arrowheads indicate where the leading edge of cones were measured in (C). (C) Cone widths across multiple cones and individualizing cysts indicate D. aur and D. pse cones are narrower than those in D. mel (D. mel: n= 38 cones across 7 testes, D. aur: n= 27 cones across 4 testes, D.pse: n= 21 cones across 7 testes).

Discussion

Arp2 arose early in eukaryotic evolution and is under stringent sequence conservation for its many roles throughout all tissue types^17,35. Our studies reveal recurrent Arp2 gene duplication and subsequent sequence diversification in Drosophila. The exclusive expression of Arp2 paralogs in the male germline suggests a selective pressure to have a testis-specific Arp2. Despite their sequence diversification, the testis paralogs can replace Arp2 in somatic tissue and lead to viable flies, even in a non-native species. However, their divergence in sequence has led to incompatibility with D. melanogaster sperm development, suggesting functional specialization for actin branching in their native species.

Gene duplications often relieve tension between conflicting roles of the parental gene^36–39. We find that Arp2 may have conflicting roles between its ‘housekeeping’ functions in somatic tissue and those in sperm development, and the optimal Arp2 sequence for somatic roles may result in a germline fitness cost. This tension can be relieved with a germline-specific Arp2 paralog while retaining canonical Arp2 for somatic roles. Arp2 appears to be the only member of the Arp2/3 complex that has repeatedly duplicated. We see one additional instance of Arp2/3 diversification in Drosophila: two isoforms of Arpc3 (‘Arpc3a’ and ‘Arpc3b’). The duplicate Arpc3b appears exclusively expressed in the ovary and localizes to actin rings found only in the female germline⁹. While additional duplicates may remain uncovered, the diversification of multiple Drosophila Arp2/3 subunits indicates that each complex member may differentially tune actin polymerization for specific germline roles, and Arp2 appears uniquely central for testis specialization.

Interestingly, despite their distinct evolutionary paths, both Arp2 paralogs lead to defects in individualization and narrower cones in a non-native species, suggesting specialization for the unique cones in their native species (Figure 7). While future work will directly test the roles of Arp2D and Arp2D2 in obscura and montium individualization, the defects produced by the paralogs in D. mel sperm development suggest that their sequence diversification has altered Arp2/3 germline activity. The kinetics of Arp2/3 complex assembly or actin polymerization may be changed. Alternatively, the paralogs may vary the stability of actin networks, yet it is unknown how such changes in actin branching would modulate cone structure and motility. Precedence exists for diversification of Arp2/3 complexes leading to an impact on polymerization rate (gene duplicates of human ArpC1 and ArpC5¹⁸) or network stability (gene duplicate of human Arp3¹⁹). These differences may not be attributed directly to the complex but may result from the loss or gain of protein interactions¹⁷. Arp2D or Arp2D2 may require a testis regulatory factor or another species-specific paralog of an Arp2/3 subunit.

Focusing on the more diverged paralog Arp2D, we discovered that divergence in the D-loop contributes to the observed testis defects in D. melanogaster. In canonical Arp2/3, the D-loop interacts with Arpc3 only in the active conformation^14,15 and is critical for Arp2/3 activation¹⁵. Drosophila’s Arpc3 isoforms are not 100% identical in sequence among D. mel and the obscura and montium clades. Therefore, the diverged D-loop of Arp2D and Arp2D2 may be incompatible with D. mel Arpc3 isoforms in the germline. Interestingly, canonical Arp2 in most eukaryotes undergoes alternative splicing, resulting in two splice variants differing only by five amino acids within the D-loop^40,41 (Figure S5B). Although the distinct functions of Arp2 splice isoforms are unknown in any species, we speculate that modulation of the D-loop can tune actin polymerization for different contexts.

The Arp2D C-terminus is also incompatible with Arp2 roles in D. melanogaster sperm development. Structural work shows that the canonical Arp2 C-terminus contacts actin¹⁴, yet less is known about its role in actin polymerization and how altering its sequence impacts function. The C-terminal tail of Arp3 serves an autoinhibitory role by binding the hydrophobic groove between subdomains 1 and 3, keeping the Arp2/3 complex in an inactive conformation until ATP⁴² or a nucleation-promoting factor⁴³ binds. Analogously, the Arp2D C-terminus may be inhibitory, yet this domain is highly charged and unlikely to bind the hydrophobic groove. Instead, the C-terminus may alter charge-charge interactions with regulators critical for actin polymerization in D. melanogaster individualization.

Despite its negative impact on canonical Arp2 in sperm development, the unique Arp2D C-terminus is required for somatic function of Arp2D in D. melanogaster. The N-terminal domain composing the majority of Arp2D appears structurally like Arp2 based on homology predictions; thus, we expected this domain would be sufficient to interact with actin and the rest of the Arp2/3 complex for actin polymerization. However, because Arp2D requires the unique C-terminus for somatic function, we infer that the C-terminus has co-evolved with the ‘Arp2-like’ N-terminal domain of Arp2D and is now integral for full-length protein activity, yet incompatible with the evolutionarily older Arp2 sequence for testis function. The C-terminus may stimulate Arp2D for nucleation activity or include a regulator binding site. Interestingly, the C-termini of Arp2D orthologs vary in length and overall charge (Figure S5B). We speculate that each ortholog’s C-terminus has co-evolved with its respective ‘Arp2-like’ domain, which has led to species-specialized germline function. Overall, the unique evolutionary diversification of Arp2 paralogs offers insight into how selective pressure acts upon the Arp2 sequence and gives rise to tissue specialization.

STAR Methods

RESOURCE AVAILABILITY

Lead Contact

Requests for further information, detailed protocols, or resources and reagents should be directed to and will be fulfilled by the lead contact, Courtney Schroeder (CourtneyM.Schroeder@utsouthwestern.edu).

Materials Availability

All Drosophila stocks and plasmids generated in the study are available upon request from the lead contact.

Data and Code Availability

• Sequences obtained in this study have been deposited at NCBI GenBank and are publicly available as of the date of publication (see also Data S1A). Accession numbers are listed in Table S2. This paper also analyzes existing, publicly available data; the accession numbers for the datasets are listed in the key resources table and Table S3.

• This paper does not report original code.

• Any additional information concerning this study is available from the lead contact upon request.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Chicken polyclonal anti-GFP	Abcam	Cat. # ab13970; RRID: AB 300798
Mouse monoclonal anti-α tubulin (DM1A)	Santa Cruz Biotechnology	Cat. # Sc-32293; RRID: AB 628412
Rabbit monoclonal anti-cleaved caspase-3 (Asp175)	Cell Signaling	Cat. # 9664S; RRID: RRID: AB 2070042
Goat anti-rabbit highly cross-adsorbed secondary antibody Alexa Fluor 488	Invitrogen	Cat. # A11034; RRID: AB_2576217
Goat anti-chicken cross-adsorbed secondary antibody Alexa Fluor Plus 488	Thermo Scientific	Cat. # A32931; RRID: AB_2762843
Goat anti-chicken secondary antibody Alexa Fluor 647	Invitrogen	Cat. # A-21449; RRID: AB_2535866
Goat anti-mouse cross-adsorbed secondary antibody Alexa Fluor 568	Thermo Fisher	Cat. # A11004; RRID: AB_2534072
Bacterial strains
NEB 5alpha	NEB	Cat. # C2987I
Chemicals, peptides, and recombinant proteins
Hoechst 33342	Invitrogen	Cat. # H3570
Trizol	Invitrogen	Cat. # 15 596 026
ProLong Diamond Antifade Mountant	Thermo Fisher	Cat. # P36970
TritonX-100	Sigma Aldrich	Cat. # T8787
Tween-20	Thermo Fisher	Cat. # AAJ20605AP
SiR-Actin	Cytoskeleton	Cat. # CY-SC001
Paraformaldehyde	Thermo Fisher	Cat. # 28908
Yeast extract	Sigma Aldrich	Cat. # Y1625
Phusion	NEB	Cat. # M0530L
Gibson Assembly Master Mix	NEB	Cat. # E2611L
DpnI	NEB	Cat. # R0176S
T4 Polynucleotide kinase	NEB	Cat. # M0201S
T4 DNA ligase	NEB	Cat. # M0202S
Critical commercial assays
ZymoPURE™ Plasmid Purification Kit, Miniprep	Zymo Research	Cat. # D4212
RNA Cleanup & Concentrator Kit	Zymo Research	Cat. # R1014
DNA clean & Concentrator Kit	Zymo Research	Cat. # D4003
SuperScript III First-Strand Synthesis	Invitrogen	Cat. # 18 080 051
Invitrogen Zero Blunt TOPO PCR Cloning Kit	Invitrogen	Cat. # 450245
NucleoBond® Xtra Midi	Takara Bio	Cat. # 740410.5
Deposited data
Arp2D2 sequences (Data S1A)	This study	GenBank Accession Codes in Table S2
Experimental models: Organisms/strains
D. melanogaster Oregon-R	Harmit S. Malik’s lab	N/A
D. melanogaster w 1118	Harmit S. Malik’s lab	N/A
Drosophila barbarae	Cornell Stock Center	14028-0491.01
Drosophila mayri	Cornell Stock Center	14028-0591.01
Drosophila birchii	Cornell Stock Center	14028-0521.00
Drosophila bicornuta	Cornell Stock Center	14028-0511.01
Drosophila seguyi	Cornell Stock Center	14028-0671.02
Drosophila nikananu	Cornell Stock Center	14028-0601.01
Drosophila diplacantha	Cornell Stock Center	14028-0586.00
Drosophila burlai	Cornell Stock Center	14028-0781.00
Drosophila vulcana	Cornell Stock Center	14028-0711.00
Drosophila punjabiensis	Cornell Stock Center	14028-0641.00
Drosophila watanabei	Cornell Stock Center	14028-0531.02
Drosophila serrata	Cornell Stock Center	14028-0681.00
Drosophila bocki	Cornell Stock Center	14028-0751.00
Drosophila kikkawai	Cornell Stock Center	14028-0561.00
Drosophila kanapiae	Cornell Stock Center	14028–0541.00
Drosophila auraria	EHIME	E-11215
Drosophila auraria	Cornell Stock Center	14028-0471.00
Drosophila triauraria	Cornell Stock Center	14028-0651.00
Drosophila rufa	Cornell Stock Center	14028-0661.02
Drosophila pseudoobscura	Cornell Stock Center	14011-0121.34
Drosophila auraria	EHIME	E-11201
Drosophila auraria	EHIME	E-11204
Drosophila auraria	EHIME	E-11206
Drosophila auraria	EHIME	E-11207
Drosophila auraria	EHIME	E-11208
Drosophila auraria	EHIME	E-11212
Drosophila auraria	EHIME	E-11213
Drosophila auraria	EHIME	E-11214
Drosophila auraria	EHIME	E-11217
Drosophila rufa	EHIME	E-14801
Drosophila rufa	EHIME	E-14805
Drosophila rufa	EHIME	E-14809
Drosophila rufa	EHIME	E-14812
Drosophila rufa	Cornell Stock Center	14028-0661.02
D. melanogaster y[1] w[1118]; PBac{y[+]-attP-9A}VK00027	Bloomington Drosophila Stock Center	Stock # 9744
D. melanogaster w[1118]; PBac{y[+mDint2] GFP[E.3xP3]=vas-Cas9}VK00027	Bloomington Drosophila Stock Center	Stock # 51324
D. melanogaster w[1118] attp 3xP3-DsRed Arp2−/FM7	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2promoter-D. mel Arp2 3xP3-DsRed Arp2−/FM7	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2promoter-D. pseudoobscura Arp2D 3xP3-DsRed Arp2−/FM7	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2promoter-D. auraria Arp2D2 3xP3-DsRed Arp2−/FM7	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2promoter-D. pseudoobscura Arp2Dcodon optimized 3xP3-DsRed Arp2−/FM7	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2 promoter-sfGFP-D. pseudoobscura Arp2	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2 promoter-sfGFP-D. pseudoobscura Arp2D	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2 promoter-sfGFP-D. auraria Arp2D2	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp53D promoter-sfGFP-D. pseudoobscura Arp2	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp53D promoter-sfGFP-D. pseudoobscura Arp2D	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2promoter-D. pseudoobscura Arp2D-2D2 loop 3xP3-DsRed Arp2−/FM7	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2promoter-D. mel Arp2-2D loop 3xP3-DsRed Arp2−/FM7	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2promoter-D. pseudoobscura Arp2D-2CT 3xP3-DsRed Arp2−/FM7	This study	N/A
D. melanogaster w[1118] 3xP3-sfGFP D.mel Arp2promoter-D. melanogaster Arp2-2DCT 3xP3-DsRed Arp2−/FM7	This study	N/A
Recombinant DNA
pDsRed-attP	Addgene	plasmid 51019
pCFD4-U6:1_U6:3tandemgRNAs	Addgene	plasmid 49411
pCFD4-Arp2 gRNA1 gRNA2	This study	N/A
pDsRed-attP-1kb 5’ and 3’ homology arms of Arp2	This study	N/A
p-attB-3xP3-sfGFP-Arp2 promoter-sfGFP-D. melanogaster Arp2	This study	N/A
p-attB-3xP3-sfGFP-Arp2 promoter-sfGFP-D. pseudoobscura Arp2D	This study	N/A
p-attB-3xP3-sfGFP-Arp2 promoter-sfGFP-D. auraria Arp2D2	This study	N/A
p-attB-3xP3-sfGFP-Arp53D promoter-sfGFP-D. melanogaster Arp2	This study	N/A
p-attB-3xP3-sfGFP-Arp53D promoter-sfGFP-D. pseudoobscura Arp2D	This study	N/A
p-attB-3xP3-sfGFP-D. melanogaster Arp2 promoter- D. melanogaster Arp2	This study	N/A
p-attB-3xP3-sfGFP-D. melanogaster Arp2 promoter- D. pseudoobscura Arp2D	This study	N/A
p-attB-3xP3-sfGFP-D. melanogaster Arp2 promoter- D. auraria Arp2D2	This study	N/A
p-attB-3xP3-sfGFP-D. melanogaster Arp2 promoter- D. pseudoobscura Arp2D codon optimized	This study	N/A
p-attB-3xP3-sfGFP D. melanogaster Arp2 promoter-D. pseudoobscura Arp2D-2D2 loop	This study	N/A
p-attB-3xP3-sfGFP D. melanogaster Arp2 promoter-D. melanogaster Arp2-2D loop	This study	N/A
p-attB-3xP3-sfGFP D. melanogaster Arp2 promoter-D. pseudoobscura Arp2D-2CT	This study	N/A
p-attB-3xP3-sfGFP D. melanogaster Arp2 promoter-D. melanogaster Arp2-2DCT	This study	N/A
Software and algorithms
ImageJ (Fiji)	Schindelin et al.64,65	RRID:SCR 002285
NIS-Elements Software	Nikon	RRID:SCR 014329
Nikon denoise.ai algorithm	Nikon; Davis, M.63	N/A
Geneious Software	Kearse et al.48	RRID:SCR 010519
Jalview	Waterhouse et al.53	RRID:SCR 006459
JPred4	Drozdetskiy et al.54	http://www.compbio.dundee.ac.uk/jpred4; RRID:SCR 016504
UCSF Chimera Software	Pettersen et al.52	RRID:SCR 004097
SWISS-MODEL	Waterhouse et al.50	RRID:SCR 018123
McDonald-Kreitman online resource	Egea et al.56	http://mkt.uab.es/mkt/help_mkt.asp
PAML suite	Yang Z.57	RRID:SCR_014932
Other
Drosophila montium Species Group Genomes Project	Bronski et al.24	BioProject ID 554346
NuPAGE LDS sample buffer	Thermo Fisher	Cat. # NP0007
NuPAGE 4–12% Bis-Tris gel	Thermo Fisher	Cat. # NP0321BOX
MES SDS running buffer	Thermo Fisher	Cat. # NP0002
Iblot™ 2 Transfer Stacks, PVDF, mini	Thermo Fisher	Cat. # IB24002
Phosphate-Buffered Saline	Sigma Aldrich	Cat. # D8537
Tris Buffered Saline,10 x, solution	Sigma Aldrich	Cat. # T5912-1L

EXPERIMENTAL MODEL AND SUBJECT DETAILS:

All flies were cultured on yeast-cornmeal-molasses-malt extract medium at room temperature, and all transgenic lines, wildtype species, and strains used in this study are listed in the key resource table. Oregon-R flies were used as wildtype flies in fertility assays, and for all crosses, female virgins that were 1–5 days old and male virgins that were 1–3 days old were used. Virgins were maintained at room temperature until crosses were set up. All crosses were set up with females in excess of males at a ratio of 5:2. Pairs were mated for approximately one week with vials being flipped every 3 days. For heat stress, the mating pairs were maintained at 29°C, and all crosses were carried out with consistent light/dark cycles. Adult progeny were quantified for no more than 16 days relative to when crosses were setup or 13 days (crosses at 29°C) to avoid counting progeny from the next generation. Parental flies that died during the mating week were tallied and did not differ significantly among genotypes. For comparing germline defects, virgin males were aged for 0–3 days before conducting testis dissections. To compare actin cones across different wildtype species, males were obtained from the following cultured stocks: Oregon-R for D. melanogaster, Cornell stock 14011–0121.34 for D. pseudoobscura, and Cornell stock 14028–0471.00 for D. auraria.

METHOD DETAILS:

Sequencing the Arp2D2 locus

Arp2D2 was initially identified by using a tBLASTn search with canonical D. melanogaster Arp2, and mRNA encoding an ‘Arp2-like’ gene was found in D. serrata (Accession XM_020944101.1) and D. kikkawai (Accession XM_017179971.1). The coding sequence was then mapped to the D. kikkawai²² and D. serrata²³ genomes (contig 7911 and MTTC01000122.1, respectively) to identify the loci and found they were syntenic and differed in canonical Arp2’s locus. For sequencing the shared syntenic locus of montium species, D. melanogaster and D. kikkawai genomes were aligned, and primers were designed to target conservation in the intergenic regions flanking the D. kikkawai Arp2D2 locus. Primers were then used for targeted sequencing of the Arp2D2 locus in unsequenced montium species. Some montium species had genomic DNA purified for a past study⁴⁶ and were kindly offered for the study. Cultured species for which genomic DNA was not already available were homogenized and DNA was extracted as done previously⁴⁶. PCRs were conducted following a touchdown protocol⁴⁷ with Phusion, following the manufacturer’s instructions (NEB). PCRs were directly sequenced, or first TOPO cloned (Invitrogen) followed by sequencing. Primers were iteratively designed using sequences from successful PCRs to obtain Arp2D2 loci sequences that were more diverged in the intergenic regions. Primer sequences are listed in Table S1, and Arp2D2 sequences obtained by PCR are provided in Data S1.

Sequence analysis with published genomes

D. melanogaster Arp2 and D. kikkawai Arp2D2 protein sequences were used in a tBLASTn search to obtain orthologs in sequenced montium species²⁴. The search set was BioProject ID 554346 (a whole genome shotgun contigs database)²⁴. Hits that had E values of 0 were obtained and syntenic orthologs of D. kikkawai Arp2D2 were verified by conservation of upstream and downstream genetic regions, including neighbors Hmu and CG3339. Arp2 orthologs were also verified by conservation of upstream and downstream regions, including genetic neighbors like Pp2B-14D and CG9903, and were consistently intronic, unlike Arp2D2. Arp2D2 coding sequences were then extracted based on alignments with D. kikkawai Arp2D2, and Arp2 coding sequences were obtained by aligning with D. kikkawai Arp2 or Arp2 from a more closely related montium species. All analyses were done using the Geneious software package⁴⁸. Accession numbers of scaffolds encoding Arp2D2 or Arp2 in BioProject 554346²⁴ are provided in Table S3.

Phylogenetics and structural analysis

Nucleotide sequences were aligned using translation alignment in the Geneious software⁴⁸. A maximum-likelihood tree was generated with PhyML (HKY85 substitution model), and 100 bootstrap replicates were performed for statistical support in Geneious⁴⁸ (see Data S1B). Protein sequence alignments were conducted using MUSCLE⁴⁹ in Geneious⁴⁸. For structural analyses, a homology model of D. pseudoobscura Arp2 was obtained using SWISS-MODEL⁵⁰, which used PDB 4JD2⁵¹ as a template, and structures were viewed using Chimera⁵². Jalview⁵³ and JPred4⁵⁴ were used for protein secondary structure prediction.

Positive selection analyses

Targeted sequencing of Arp2D2 was done with multiple strains of D. auraria and D. rufa (Table S2) using species-specific primers (Table S1). The Arp2D2 coding sequences were aligned using MUSCLE alignment in the software Geneious⁴⁸ and gaps were removed. Aligned sequences were used in an unpolarized McDonald-Kreitman test⁵⁵ using an online resource⁵⁶. We also assessed site-specific positive selection using a codon-based alignment in Geneious⁴⁸ of Arp2D2 coding sequences from 17 species (Table S2). A species tree was generated from the alignment in Geneious⁴⁸, and both the tree and sequence alignment were in the CODEML algorithm in the PAML suite⁵⁷. The program determines if the Arp2D2’s evolution fits the NSsites models M7 and M8a, which do not allow for positive selection, or the M8 model, which allows for positive selection. We used the starting omega of 0.4 and F3×4 codon frequency table. Statistical significance of the difference between the models’ log-likelihoods was assessed using a chi-squared test.

Generation of fly transgenics:

Generation of GFP-tagged Arp2 and paralogs

To visualize Arp2 variants, D. pseudoobscura Arp2 (intronless, short splice variant), D. pseudoobscura Arp2D, and D. auraria Arp2D2 were C-terminally tagged with superfolder GFP²⁶ and cloned into a vector encoding attB. Each construct was cloned with either the promoter for D. melanogaster Arp2 or for D. melanogaster Arp53D (a gene highly expressed in the testis). Constructs were midi-prepped (Takara Bio), and for site-directed transgenesis, constructs were co-injected with PhiC31⁵⁸ into a fly line encoding attP on the third chromosome (stock 9744 from Bloomington Drosophila Stock Center) by Rainbow Transgenic Flies, Inc. See construct sequences in Data S1C.

Generation of Arp2 Knockout

Arp2 was knocked out using CRISPR/Cas9 and replaced with DsRed to track the knockout allele. The guide RNAs were chosen based on a low probability of off-targets (http://flyrnai.org/crispr2/) and targeted in the intergenic regions upstream and downstream of the coding regions to completely remove the gene. The guide RNAs (GTAGCTGCTACTAGCAGACT and TCGTTACTCCCCAGAGTTGA) were cloned into pCFD4 (Addgene plasmid 49411)⁵⁹ and the 1kb regions upstream and downstream of the cut sites were cloned into vector pDsRed-attP (Addgene plasmid 51019)⁶⁰. Plasmids were midi-prepped (Takara Bio), and BestGene, Inc. injected them into a nanos-Cas9 expressing fly line (51324 from Bloomington Drosophila Stock Center). G0 adults were crossed to w¹¹¹⁸ flies to identify transformants (DsRed-fluorescent eyes) and eliminate Cas9. Transformants’ X-chromosome was balanced with FM7. We sequence verified CRISPR/Cas9 cut sites and presence of DsRed. The KO allele was only found in heterozygous females, as expected for the Arp2-KO phenotype, and no homozygous (or DsRed-positive) males were identified.

Generation of transgenic flies with Arp2-KO

Several transgenic lines were generated with the Arp2-knockout fly line to express the Arp2 paralogs or chimeras in the absence of wildtype, endogenous Arp2. To clone the constructs for transgenesis, the genes encoding D. pseudoobscura Arp2D and D. auraria Arp2D2 were obtained by PCR from the species’ genomic DNA. Canonical Arp2 was obtained by PCR from D. melanogaster cDNA, and the short splice variant (lacking alternative exon 3) was used for this study. Codon-optimized Arp2D was synthesized by Integrated DNA Technologies, Inc. All mutants (Arp2D-2D2 loop, Arp2-2D loop, Arp2D-2CT, and Arp2-2DCT) were generated using PCR site-directed mutagenesis, followed by DpnI treatment, polynucleotide kinase phosphorylation, and T4 ligation, following manufacturer’s instructions (NEB). All genes were Gibson cloned (NEB) into a vector encoding an attB site and superfolder GFP under the control of an eye-specific promoter (3XP3), which was used as a marker for identifying genotypes. Flanking all genes were approximately the 1kb upstream and downstream regions of canonical D. melanogaster Arp2 (see construct sequences in Data S1C). All constructs were midi-prepped and injected into the attP-encoding Arp2 KOs. The Arp2D-2D2 loop, Arp2D-2CT, and Arp2-2DCT constructs were injected by Rainbow Transgenic Flies, Inc, and for generating the Arp2-2D loop transgenic, we co-injected the construct with PhiC31 integrase following standard methods⁶¹. We crossed G0 adults to Arp2-KO stocks, selected flies with GFP- and DsRed-positive eyes and balanced the X-chromosome with FM7. To sequence verify all transgenics, genomic DNA from flies was obtained by grinding two flies in 10 mM Tris-HCl pH8, 1 mM EDTA, 25 mM NaCl, and 200 μg/mL Proteinase K. The lysate was incubated at 37°C for 30 minutes then at 95°C for 3 minutes to deactivate Proteinase K. The lysate was briefly centrifuged, and the supernatant was used for PCR-amplification of the modified loci. A touchdown protocol⁴⁷ was conducted with Phusion (NEB), and PCR products were used for Sanger sequencing.

RT-PCR

To assess transgene expression, whole flies (females versus males) were used for RNA extractions. To detect expression in the germline versus the carcass, abdomens were separated from the body and further dissected in PBS to extract testes or ovaries; the head was further separated from the remaining carcass in Figure 1D to localize Arp2D2 expression in the germline, head, or remaining tissue. Tissues or whole flies (5–10 total) were homogenized with TRIzol (Invitrogen) then briefly centrifuged to separate the supernatant. The supernatant was chloroform-extracted, resulting in a soluble phase that was then isopropanol-extracted. The precipitated RNA was briefly centrifuged, and the resulting pellet was washed with 70% ethanol and resuspended in RNAse-free water. Samples were treated with DNase I (Zymo Research), followed by purification and concentration (RNA Cleanup & Concentrator kit, Zymo Research). The samples were used to generate cDNA using SuperScript III First-Strand Synthesis, following the manufacturer’s instructions (Invitrogen). The cDNA was then used to conduct reverse-transcription PCRs with Phusion (NEB). Primers are provided in Table S1.

Immunoblot analysis

Approximately 30 ovaries from the D. mel line w¹¹¹⁸ (control) and transgenic lines sfGFP-Arp2, sfGFP-Arp2D, and sfGFP-Arp2D2 were dissected in PBS and flash frozen. The pellets were resuspended in 50 μL of 4x NuPAGE LDS sample buffer (Thermo Fisher) and boiled at 95°C for 5 minutes. Protein samples were loaded on a NuPAGE 4–12% Bis-Tris gel (Thermo Fisher), run with MES SDS buffer, and then transferred to a PVDF membrane (Thermo Fisher). The membrane was blocked with 5% milk in Tris-buffered saline and 0.1% Tween-20 (TBST) then probed with anti-GFP diluted 1:500 (Abcam) and anti-tubulin diluted 1:250 (Santa Cruz Biotechnology) in TBT overnight at 4°C. The membrane was washed in TBST three times for 10 minutes then incubated with AlexaFluor 647 anti-chicken and AlexaFluor 568 anti-mouse diluted 1:3,000 (Thermo Fisher) for 1 hour at room temperature. Following three washes, the blot was scanned at 647 nm and 546 nm.

Microscopy:

Ovary imaging

Virgin females were fed yeast paste (Sigma Aldrich) for two days prior to ovary dissections, and ovaries were dissected in PBS. For live imaging of ovaries expressing GFP-tagged Arp2, Arp2D or Arp2D2, ovaries were incubated with Hoechst 33342 (Invitrogen) diluted in PBS for 10 minutes at room temperature and then imaged. For imaging Arp2-knockout flies that express Arp2 or Arp2D (untagged), ovaries were fixed in 4% paraformaldehyde (PFA) for 10 min at room temperature. Following washes with PBS, fixed ovaries were incubated with 2 μM SiR-Actin⁴⁵ (Cytoskeleton, Inc.) overnight at 4°C. Ovaries were then washed with PBS, stained with Hoechst for 10 min at room temperature, and mounted on slides with ProLong Diamond Antifade Mountant (Thermo Fisher). All ovary images (live and fixed samples) were collected on the Nikon CSU-W1 with SoRa (UTSW’s Quantitative Light Microscopy Core).

Testis imaging

For brightfield imaging of testes or seminal vesicles, similarly aged males were collected and dissected in PBS. They were then placed on slides for live imaging. Images were collected on a Nikon Ts2 epi-fluorescent microscope.

For confocal imaging, virgin male flies were dissected at 0–3 days old. For fixed whole mount testes and/or seminal vesicles, germline tissue was dissected in PBS and then transferred to an eppendorf tube for fixation using 4% PFA (25 min at room temperature). Samples were then washed twice for 15 min each with PBS including 0.1% Triton. To probe for actin and DNA, testes were stained with 2 μM SiR-Actin⁴⁵ (Cytoskeleton, Inc.) for 2–3 hours at room temperature and incubated with Hoechst 33342 (Invitrogen) for 10 min at room temperature. Stained testes were then placed on a slide, non-germline tissue was further removed, and a coverslip was placed following addition of ProLong Diamond Antifade Mountant (Thermo Fisher). For D. pse testes, testes had to be punctured to release developing cysts, as testes were too thick to image unlike D. mel and D. aur testes, and fixed on the slide as done previously⁶². To probe for GFP in D. mel testes, whole mount testes were permeabilized with 0.3% TritonX-100 for 30 min following fixation. After washing with 0.1% Triton, testes were incubated with anti-GFP (Abcam ab13970) at 1:500 in blocking solution (3% bovine serum albumin in PBS) overnight at 4°C. Testes were then washed and incubated with anti-chicken Alexa Fluor Plus 488 (Thermo Scientific at 1:2,500) and 2 μM SiR-Actin⁴⁵ (Cytoskeleton, Inc.) for 2 hrs at room temperature. Testes were washed, stained with Hoechst 33342 (Invitrogen), and mounted as done with other testis samples. To probe for activated caspase 3, whole mount testis imaging was conducted similarly, though testes were stained with anti-cleaved caspase 3 (Asp 175, Cell Signaling) as done previously³³ and probed with anti-rabbit secondary antibody (Alexa Fluor 488, Invitrogen). All testis imaging was done using an inverted Ti2 Nikon AX-R confocal microscope and NIS-Elements (Nikon), and Nikon’s denoise.ai algorithm⁶³ was used to remove background noise.

QUANTIFICATION AND STATISTICAL ANALYSIS

Male germline defects

Z-stacks were acquired of whole mount testes that were stained for actin and DNA. Stacks were then visualized with NIS-Elements (Nikon) to quantify defects (at room temperature), including cone widths, asynchronous cones, misoriented cones, and waste bags. Cones that were no longer localized to sperm nuclei were classified as motile and quantified as asynchronous if a cyst had more than three cones that were trailing behind most cones. Cones that were not trailing behind most other cones in a cyst and were mid-individualization were measured at the leading edge using the line tool in Fiji^64,65 to compare cone widths among different genotypes. Only cones that were parallel to the plane of the image and not obscured by other cones were measured. Waste bags were identified by degrading, amorphous actin cones. Misoriented cones were identified by the front of actin cones facing the basal end instead of the apical end. When counts of waste bags, counts of misoriented cones, or cone widths were compared, the statistical significance was evaluated with a t-test. When the percentage of asynchronous cones and testes with abnormal morphology were compared, statistical significance was assessed using a chi-squared test, using values for Arp2-expressing flies as the expectation. Seminal vesicle size was measured using the line tool in Fiji^64,65, and a t-test was used to evaluate statistical significance. A minimum of 10 flies were dissected for each analysis, and specific sample sizes for analyses, including the number of testes or cones quantified, are noted in the figure legends. Immunofluorescence images in all figures are maximum projections of z-stack slices (unless noted otherwise) to show the entire cone, cyst, or testis (excluding the epithelium) where needed.

Fly crosses

All fertility assays were done at least three times, and statistical significance was evaluated with a t-test. For assays testing for rescue of Arp2-KO lethality, at least 5 replicates were done (noted in legends), and genotypes were scored per vial. Averages with variance are displayed. To assess statistical significance, the genotypes were summed across replicates and the observed ratio of genotypes was compared to the expected Mendelian proportion with a chi-squared test. A p-value less than 0.05 resulted if observed ratios deviated significantly from the Mendelian expectation.

Supplementary Material

Data S1

Data S1: Data supporting the key findings, related to Figures 1–6. Table A lists the Arp2D2 sequences obtained in this study in addition to Arp2 and Arp2D2 coding sequences from BioProject ID 554346²⁴ and annotated genomes^22,23,44. The strain sequences of D. auraria and D. rufa Arp2D2 (used in the McDonald-Kreitman test for positive selection) are included and annotated with the stock number of the corresponding strain (also listed in Table S2). GenBank accession codes for sequences obtained in this study are included in the sequence headings. Table B includes the tree for Figure 1B as a newick file with bootstrap support. Table C lists the construct sequences used for generating fly transgenics.

Highlights.

• Arp2 has undergone recurrent, species-specific diversification in Drosophila

• Testis-enriched Arp2 paralogs are incompatible for cross-species spermatogenesis

• Sequence divergence in the D-loop and C-terminus contribute to incompatibility

• The unique C-terminus of one paralog is critical for its function

Inclusion and diversity

We support inclusive, diverse, and equitable conduct of research.