**1. Introduction**

The P-element-induced wimpy testis (PIWI)-interacting RNA (piRNA) pathway is an RNA interference (RNAi) mechanism that is traditionally known to silence transposable elements (TEs) that can integrate into the germline genome and threaten its integrity [1–4]. piRNAs, 23–30 nucleotide (nt) small RNAs (sRNAs), bind Piwi proteins, a subfamily of the Argonautes. piRNA-bound Piwis assemble into piRNA-induced silencing complexes (piRISCs), where effector Piwis are targeted to complementary RNA substrates [5–7]. *Drosophila* express three Piwis (Piwi, Aubergine (Aub), and Argonaute-3 (Ago3)), where Aub and Ago3 are expressed exclusively in the germline, while Piwi is also expressed in neighboring somatic cells [8–10]. Aub and Ago3 silence their targets post-transcriptionally in the cytoplasm, while Piwi translocates into the nucleus, forms a nuclear effector complex, and silences its targets co-transcriptionally [11–13]. Depletion of the piRNA pathway in *Drosophila* leads to TE insertion accumulation, DNA damage, defects in embryonic development, and female sterility [14–19].

**Citation:** Williams, A.E.; Shrivastava, G.; Gittis, A.G.; Ganesan, S.; Martin-Martin, I.; Valenzuela Leon, P.C.; Olson, K.E.; Calvo, E. *Aedes aegypti* Piwi4 Structural Features Are Necessary for RNA Binding and Nuclear Localization. *Int. J. Mol. Sci.* **2021**, *22*, 12733. https://doi.org/10.3390/ ijms222312733

Academic Editor: Michail Kotsyfakis

Received: 31 October 2021 Accepted: 23 November 2021 Published: 25 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

The functions of the piRNA pathway are more diverse than initially thought [3,20–23]. For example, Piwis display differential expression patterns in the germline or soma as well as in the cytoplasm or the nucleus across organisms [21], suggesting that their roles and functions are broad. A good example is in arthropods, where both somatic and germline piRNAs are common [24]. Culicine mosquitoes, specifically, have undergone an expansion of their Piwi protein repertoire, suggesting functional divergence [25]. For example, *Aedes* express seven Piwis, Piwi2-7, and Ago3, where Piwi4-6 and Ago3 are abundantly expressed in both the soma and germline [25,26]. Many studies have shown that the piRNA pathway in *Aedes* is multi-functional and important for, in addition to transposon silencing, antiviral immunity [25,27–38], embryonic development [39,40], and gene regulation [41,42].

A particularly interesting *A. aegypti* Piwi of unknown function is Piwi4 (henceforth termed "AePiwi4"). Although unnecessary for small RNA production, AePiwi4 associated with Semliki Forest virus-specific small-interfering RNAs (vsiRNAs) and virus-specific piRNAs (vpiRNAs) in infected cells, as well as with several protein players involved in both the siRNA (Ago2 and Dcr2) and piRNA (Ago3, Piwi5, Piwi6, Yb, vreteno, Tejas, and minotaur) pathways [26,43]. Furthermore, silencing *AePiwi4* depleted 30 2 0 O-methylated (mature) Sindbis virus (SINV)-specific vsiRNAs and vpiRNAs and increased SINV, dengue (DENV2), and chikungunya (CHIKV) virus replication in infected cells [32]. This phenotype was recapitulated in DENV2-infected *A. aegypti* mosquitoes, where silencing *AePiwi4* increased infectious virus titers 5–10 days post-infection (dpi) [32]. AePiwi4 also associated with highly conserved satellite repeat-derived piRNAs (tapiR1 and tapiR2) that were 3 0 2 0 O-methylated [39]. Knocking down *AePiwi4* reduced tapiR1 and tapiR2 transcripts, and depleting tapiR1 in embryos arrested their development and prevented the degradation of maternally deposited transcripts [39]. Taken together, the role(s) of AePiwi4 appear to be diverse and span across several different RNAi pathways.

AePiwi4 has been consistently associated with long (28–30 nt), mature 30 2 0 O-methylated (henceforth termed "30m") piRNAs, and it was found in both the cytoplasmic and nuclear fractions in an embryonic *Aedes aegypti* cell line (Aag2) [26,32,39]. We therefore set out to characterize AePiwi4 structural motifs involved in piRNA binding and nuclear localization to gain further insights on AePiwi4 function. In human Piwis (Hiwi1, Hiwi2, and Hili), the PAZ (Piwi/Argonaute/Zwille) domain preferentially binds 30m piRNAs because of a hydrophobic-binding pocket that is flexible enough to accommodate the methyl group [44]. This contrasts with the human Argonaute-1 (Ago1) PAZ domain where its more restrictive RNA-binding pocket exhibits preferential binding to 30 2 0 -OH (henceforth termed "30nm") groups present on microRNAs [44]. We hypothesized that a flexible AePiwi4 PAZ domain would also determine AePiwi4 preferential binding to mature, long piRNA populations.

We first compared PAZ sequences across previously crystalized Piwis to determine AePiwi4 PAZ structural features and binding pockets involved in 30 end piRNA recognition. We then characterized recombinant AePiwi4 PAZ-binding dynamics with the 30 ends of mature and un-methylated piRNAs by surface plasmon resonance (SPR). We found that mutating putative RNA-binding residues depleted or significantly impacted binding to both 30m and 30nm sRNAs, while a T41R change, present in *A. aegypti* Ago3, significantly improved binding. Finally, we characterized a functional nuclear localization signal (NLS) in the N-terminal region of the AePiwi4 protein. We found that subtle structural differences across Piwi proteins may have important impacts on preferential RNA-binding behaviors and subcellular localization.

#### **2. Results**

#### *2.1. Biophysical Properties of AePiwi4 by Structural Modeling and Alignment*

Using I-TASSER and Chimera, we first modeled AePiwi4 against the recently crystalized *Drosophila melanogaster* Piwi protein [45] (PDB: 6KR6) (Figure 1A). The quality of the predicted model was assessed by its C-score = −1.50. C-scores fall between −5 and 2, and more than 90% of the quality predictions are correct for models that have C-scores of −1.5 or higher [46]. Furthermore, the average template modeling (TM) score (0.53 ± 0.15) was >0.5, which indicates a model of correct topology [46]. We then superimposed the AePiwi4 model against crystalized human (Hiwi1; PDB: 3O7V) [44] and mouse (Miwi; PDB: 2XFM) [47] PAZ to determine the amino acid residues of AePiwi4 PAZ to be M270 –T380 (Figure 1B). A summary of predicted biophysical properties of AePiwi4 PAZ is provided in Supplemental Table S1. An electrostatic density map of Piwi4 (Figure 1C) revealed an inner pocket that was highly positively charged, analogous to the *Drosophila* Piwi linker regions that bind RNA nucleotides. AePiwi4 PAZ also displayed long stretches of flexibility with neighboring hydrophobic regions (Supplemental Figure S1). The AePiwi4 PAZ model suggests that the protein contains hydrophobic regions buried within a flexible protein structure, allowing AePiwi4 to bind long 30m piRNAs.

To determine putative AePiwi4 PAZ amino acids involved in RNA binding, we aligned Piwi PAZ sequences derived from proteins whose structures had been crystalized, including *Drosophila* Piwi [45] (PDB: 6KR6), silkworm Piwi (Siwi [48]) (PDB: 5GUH), mouse Piwi (Miwi [47]) (PDB: 2XFM), and human Hili (PDB: 3O7X) and Hiwi1 (PDB: 3O6E) [44] (Figure 1D). We also included human Hiwi2 PAZ, as its binding properties to the 30 ends of piRNAs were characterized by isothermal calorimetry in Tian et al., 2011 [44]. As a comparison, we included the outgroup Argonaute protein human Ago1 (PDB: 4KXT), whose more restricted PAZ domain dictates its RNA-binding preference for 30nm microRNAs [44,49]. We then compared known RNA-binding residues from the crystal structures with the residues of *A. aegypti* PAZ domains (black arrows, Figure 1D; black arrows, Supplemental Figure S1). We found that the residues involved in RNA binding tended to be highly conserved across the different organisms, and many were tyrosine and phenylalanine aromatic residues, whose hydroxyl groups form hydrogen bonds with phosphate oxygens of RNA nucleotides [44]. We also noted that all Piwi PAZ domains analyzed herein displayed the Piwi PAZ specific insertion element (black box, Figure 1D) that provides the flexibility necessary for accommodating 30m piRNA ends [44]. This insertion site lies between two beta barrels, which, when absent (as in Ago1), results in a sharp turn between the barrels and a narrower binding pocket [44]. Although the amino acids within the Piwi PAZ-specific insertion site are not conserved across organisms, or even across subfamily Piwi proteins in the same organism [44], we observed that the first five amino acids of the Piwi PAZ-specific insertion sites were highly conserved across the *A. aegypti* Piwis. The only exception to this observation was *A. aegypti* Ago3, whose residues shared no similarity to its AePiwi PAZ counterparts. Further, Ago3 displayed two amino acids within the insertion site not seen in the other PAZ sites, perhaps suggesting a more flexible binding pocket than the other *A. aegypti* Piwis.

We also generated a phylogenetic tree to compare evolutionary relatedness between Piwi PAZ sequences from the various organisms (Figure 1E). We included human Ago1 again as an outgroup. We also added the two additional *Drosophila* Piwis, Aub, and Ago3. We found that all *A. aegypti* Piwi PAZ, except for Ago3, clustered with *Drosophila* Aub. On the other hand, *A. aegypti* Ago3 was most closely related to *Drosophila* Ago3. We also observed AePiwi4 clustered with the germline AePiwis Piwi2-3 as opposed to the somatic AePiwis Piwi5, Piwi6, and Ago3.

**Figure 1.** Predicted *A. aegypti* Piwi4 RNA-binding properties. (**A**) Predicted *A. aegypti* Piwi4 model (blue) superimposed with the crystalized *Drosophila* Piwi structure (purple); (**B**) predicted *A. aegypti* Piwi4 structure (blue) with PAZ domain highlighted in orange; (**C**) electrostatic density of *A. aegypti* Piwi4 where red = negatively charged and blue = positively charged. Structure is rotated on the right to reveal inner positively charged pocket; (**D**) alignment of Piwi PAZ domains, including all *A. aegypti* Piwis (Ae. Piwi2-7 and Ae. Ago3) and crystalized or characterized Piwi PAZ (*Drosophila* (Dme) Piwi, silkworm Piwi (Siwi), mouse Piwi (Miwi), and human Piwis Hiwi1, Hiwi2, and Hili). The human Argonaute protein Ago1 was also included as an outgroup. Black arrows indicate known RNA-binding residues by crystal structures, which in this alignment include amino acid numbers 50, 56, 71, 88, 89, 93, 123, and 125. Black box indicates the Piwi PAZspecific insertion site. Colors (Zappo color scheme) indicate biochemical properties where peach = aliphatic/hydrophobic, aromatic = orange, blue = positively charged, red = negatively charged, green = hydrophilic, pink = conformationally special, and yellow = cysteine; (**E**) phylogenetic tree of all Piwi PAZ included in the alignment shown in 1D, with the addition of *Drosophila* Aubergine and Ago3 PAZs. Scale bar indicates number of substitutions per site.

#### *2.2. A. aegypti Piwi4 PAZ Binds 3*0 *2* 0 *O-Methylated and Non-Methylated piRNAs in a Sequence-Independent Manner* To determine whether *A. aegypti* Piwi4 PAZ bound to both 3′ 2′ O-methylated and 3′ 2′ OH piRNAs, we cloned *AePiwi4 PAZ* with a histidine 6xHis-tag into pET-17b for bacte-

*2.2. A. aegypti Piwi4 PAZ Binds 3′ 2′ O-Methylated and Non-Methylated piRNAs in a* 

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 5 of 22

*Sequence-Independent Manner* 

To determine whether *A. aegypti* Piwi4 PAZ bound to both 30 2 0 O-methylated and 3 0 2 0 OH piRNAs, we cloned *AePiwi4 PAZ* with a histidine 6xHis-tag into pET-17b for bacterial expression (Supplemental Figure S2A,B), purified it, and characterized RNAbinding dynamics by SPR. We confirmed AePiwi4 PAZ expressed in BL21(DE3) pLysS *E. coli* by Western blot using antibodies against its 6xHis-tag (Supplemental Figure S2C). We purified AePiwi4 PAZ from the soluble fraction by nickel chromatography followed by size exclusion chromatography. Soluble AePiwi4 PAZ was stable in 20 mM Tris-HCl pH 7.4 and 150 mM NaCl and ran at the expected size (14 kDa) by SDS/PAGE gel electrophoresis (Supplemental Figure S2D). Protein identity was confirmed by Edman degradation. rial expression (Supplemental Figure S2A,B), purified it, and characterized RNA-binding dynamics by SPR. We confirmed AePiwi4 PAZ expressed in BL21(DE3) pLysS *E. coli* by Western blot using antibodies against its 6xHis-tag (Supplemental Figure S2C). We purified AePiwi4 PAZ from the soluble fraction by nickel chromatography followed by size exclusion chromatography. Soluble AePiwi4 PAZ was stable in 20 mM Tris-HCl pH 7.4 and 150 mM NaCl and ran at the expected size (14 kDa) by SDS/PAGE gel electrophoresis (Supplemental Figure S2D). Protein identity was confirmed by Edman degradation. To test RNA-binding dynamics, we performed SPR with AePiwi4 PAZ and three different RNAs: a 3′m 28 nt Phasi Charoen-like virus (PCLV)-specific piRNA, a 3′nm vpiRNA

To test RNA-binding dynamics, we performed SPR with AePiwi4 PAZ and three different RNAs: a 30m 28 nt Phasi Charoen-like virus (PCLV)-specific piRNA, a 30nm vpiRNA of the same sequence, and a 30nm 28 nt scrambled sequence. We chose this piRNA sequence because a recent publication noted that PCLV piRNAs are broadly distributed across culicine mosquito cell lines, perhaps due to a PCLV-specific endogenous viral element in the genome [50]. Using synthetic RNA sequences biotinylated on the 50 end, we immobilized the RNA on a CM5 Biacore surface that had been pre-coated with neutravidin. Immobilizing RNA by the 50 end allowed us to test binding affinities to moieties at the 30 end. Using increasing concentrations of AePiwi4 PAZ analytes flowed over the surface of the chip with immobilized ligand, we were able to determine the dissociation constants (KD) from steady-state binding levels (Req) against the analyte concentration (C, in molar concentration) once binding reached equilibrium. Experiments were performed in four replicates. of the same sequence, and a 3′nm 28 nt scrambled sequence. We chose this piRNA sequence because a recent publication noted that PCLV piRNAs are broadly distributed across culicine mosquito cell lines, perhaps due to a PCLV-specific endogenous viral element in the genome [50]. Using synthetic RNA sequences biotinylated on the 5′ end, we immobilized the RNA on a CM5 Biacore surface that had been pre-coated with neutravidin. Immobilizing RNA by the 5′ end allowed us to test binding affinities to moieties at the 3′ end. Using increasing concentrations of AePiwi4 PAZ analytes flowed over the surface of the chip with immobilized ligand, we were able to determine the dissociation constants (KD) from steady-state binding levels (Req) against the analyte concentration (C, in molar concentration) once binding reached equilibrium. Experiments were performed in four replicates. We found that AePiwi4 PAZ bound the 3′m 28 nt piRNA with a KD of 1.7 ± 0.8 μM (Figure 2A), the 3′nm 28 nt piRNA with a KD of 5.0 ± 2.2 μM (Figure 2B), and the scrambled

We found that AePiwi4 PAZ bound the 30m 28 nt piRNA with a K<sup>D</sup> of 1.7 ± 0.8 µM (Figure 2A), the 30nm 28 nt piRNA with a K<sup>D</sup> of 5.0 ± 2.2 µM (Figure 2B), and the scrambled 28 nt 30nm piRNA with a K<sup>D</sup> of 2.5 ± 0.1 µM (Figure 2C). AePiwi4 PAZ bound to 3<sup>0</sup> 2 0 O-methylated piRNAs with marginally greater affinity than it did to 30 2 0 unmethylated piRNAs (*p* = 0.05), and there was no significant difference in binding affinities for known or scrambled RNA sequences (*p* = 0.25). 28 nt 3′nm piRNA with a KD of 2.5 ± 0.1 μM (Figure 2C). AePiwi4 PAZ bound to 3′ 2′ Omethylated piRNAs with marginally greater affinity than it did to 3′ 2′ unmethylated piR-NAs (*p* = 0.05), and there was no significant difference in binding affinities for known or scrambled RNA sequences (*p* = 0.25).

**Figure 2.** Affinity-binding equilibrium curves for AePiwi4 PAZ to piRNAs. Fitted affinity-binding equilibrium curves for AePiwi4 PAZ analyte to a 28 nt (**A**) 3′ 2′ O-methylated, (**B**) non-methylated, or (**C**) scrambled 28 nt non-methylated piRNA. Equilibrium KD was calculated from steady-state binding levels Req = (CRmax)/(KD + C) + offset, where C = concentration, Rmax = analyte-binding capacity of the surface in response units (RU), KD = dissociation constant, and offset = response at zero analyte concentration. M = molar concentration. Red bars indicate mean and standard deviation for Rmax values. **Figure 2.** Affinity-binding equilibrium curves for AePiwi4 PAZ to piRNAs. Fitted affinity-binding equilibrium curves for AePiwi4 PAZ analyte to a 28 nt (**A**) 30 2 0 O-methylated, (**B**) non-methylated, or (**C**) scrambled 28 nt non-methylated piRNA. Equilibrium K<sup>D</sup> was calculated from steady-state binding levels Req = (CRmax)/(K<sup>D</sup> + C) + offset, where C = concentration, Rmax = analyte-binding capacity of the surface in response units (RU), K<sup>D</sup> = dissociation constant, and offset = response at zero analyte concentration. M = molar concentration. Red bars indicate mean and standard deviation for Rmax values.

#### *2.3. A. aegypti Piwi4 PAZ Mutants Reveal the Amino Acids Necessary for piRNA Binding 2.3. A. aegypti Piwi4 PAZ Mutants Reveal the Amino Acids Necessary for piRNA Binding*

For further insights into how the AePiwi4 PAZ structure dictates its RNA-binding preferences, we generated AePiwi4 PAZ mutants that displayed amino acid changes For further insights into how the AePiwi4 PAZ structure dictates its RNA-binding preferences, we generated AePiwi4 PAZ mutants that displayed amino acid changes within predicted RNA-binding pockets. We focused our efforts around two highly conserved residues shown to form hydrogen bonds with RNA—Y40 and F55—as well as two

residues that flank Y40 and appeared to be moderately conserved across the *A. aegypti* Piwi PAZ—T39 and T41. Through site-directed mutagenesis, we generated five mutants that we then expressed in bacteria and purified: T39A, Y40A, T41A, F55A, and T41R (Supplemental Figures S3 and S4). Y40A and F55A displayed alanine substitutions for the highly conserved tyrosine or phenylalanine amino acids, respectively, while T39A and T41A displayed alanine substitutions for the threonines that flank Y40. The T41R mutation reflected the arginine present in only one *A. aegypti* Piwi, Ago3, but also in most other organisms' Piwi PAZ domains analyzed herein (Figure 1D). Inserts were confirmed by Sanger sequencing, and mutations in protein sequence were confirmed by mass spectrometry. Binding behaviors were assessed by SPR, as described previously (Figure 3). sential for both 3′m and 3′nm binding because when mutated, no binding occurred for either ligand. Y40A also depleted 3′nm binding and significantly inhibited 3′m piRNA binding (KD = 5.5 ± 0.5 μM; *p* = 0.04). We found that disrupting the amino acids flanking Y40 with alanine mutations had no significant impact on binding 3′m piRNAs as compared to wild-type PAZ binding to this RNA (T39A: *p* = 0.2; T41A: *p* = 0.3). However, we did observe a significantly increased affinity of T39A for the 3′nm piRNA (KD = 2.8 ± 0.4 μM; *p* = 0.02), suggesting that this residue does have an impact on 3′nm binding. Furthermore, we observed that mutating T41 to match the amino acid present in *A. aegypti* Ago3 PAZ tended to improve 3′m binding (KD = 0.57 ± 0.1 μM) and significantly improved 3′nm binding (KD = 2.0 ± 0.5 μM; *p* = 0.02).

Dissociation constants for all mutant proteins binding both the 3′m and 3′nm 28 nt piRNA are displayed in Figure 3 and summarized in Table 1. We found that F55 was es-

within predicted RNA-binding pockets. We focused our efforts around two highly conserved residues shown to form hydrogen bonds with RNA—Y40 and F55—as well as two residues that flank Y40 and appeared to be moderately conserved across the *A. aegypti* Piwi PAZ—T39 and T41. Through site-directed mutagenesis, we generated five mutants that we then expressed in bacteria and purified: T39A, Y40A, T41A, F55A, and T41R (Supplemental Figures S3 and S4). Y40A and F55A displayed alanine substitutions for the highly conserved tyrosine or phenylalanine amino acids, respectively, while T39A and T41A displayed alanine substitutions for the threonines that flank Y40. The T41R mutation reflected the arginine present in only one *A. aegypti* Piwi, Ago3, but also in most other organisms' Piwi PAZ domains analyzed herein (Figure 1D). Inserts were confirmed by Sanger sequencing, and mutations in protein sequence were confirmed by mass spectrometry. Binding behaviors were assessed by SPR, as described previously (Figure 3).

*Int. J. Mol. Sci.* **2021**, *22*, x FOR PEER REVIEW 6 of 22

**Figure 3.** Affinity-binding equilibrium curves for AePiwi4 PAZ mutants to piRNAs. Fitted affinity-binding equilibrium curves for AePiwi4 PAZ mutant analytes to a 28 nt (**A**) 3′ 2′ O-methylated or (**B**) non-methylated piRNA. Equilibrium KD was calculated from steady-state binding levels Req = (CRmax)/(KD + C) + offset, where C = concentration, Rmax = analytebinding capacity of the surface in response units (RU), KD = dissociation constant, and offset = response at zero analyte concentration. M = molar concentration. Red bars indicate mean and standard deviation for Rmax values. **Figure 3.** Affinity-binding equilibrium curves for AePiwi4 PAZ mutants to piRNAs. Fitted affinity-binding equilibrium curves for AePiwi4 PAZ mutant analytes to a 28 nt (**A**) 30 2 0 O-methylated or (**B**) non-methylated piRNA. Equilibrium K<sup>D</sup> was calculated from steady-state binding levels Req = (CRmax)/(K<sup>D</sup> + C) + offset, where C = concentration, Rmax = analytebinding capacity of the surface in response units (RU), K<sup>D</sup> = dissociation constant, and offset = response at zero analyte concentration. M = molar concentration. Red bars indicate mean and standard deviation for Rmax values.

Dissociation constants for all mutant proteins binding both the 30m and 30nm 28 nt piRNA are displayed in Figure 3 and summarized in Table 1. We found that F55 was essential for both 30m and 30nm binding because when mutated, no binding occurred for either ligand. Y40A also depleted 30nm binding and significantly inhibited 30m piRNA binding (K<sup>D</sup> = 5.5 ± 0.5 µM; *p* = 0.04). We found that disrupting the amino acids flanking Y40 with alanine mutations had no significant impact on binding 30m piRNAs as compared to wild-type PAZ binding to this RNA (T39A: *p* = 0.2; T41A: *p* = 0.3). However, we did observe a significantly increased affinity of T39A for the 30nm piRNA (K<sup>D</sup> = 2.8 ± 0.4 µM; *p* = 0.02), suggesting that this residue does have an impact on 30nm binding. Furthermore, we observed that mutating T41 to match the amino acid present in *A. aegypti* Ago3 PAZ tended to improve 30m binding (K<sup>D</sup> = 0.57 ± 0.1 µM) and significantly improved 30nm binding (K<sup>D</sup> = 2.0 ± 0.5 µM; *p* = 0.02).

**Table 1.** Summary of disassociation constants for AePiwi4 PAZ mutants binding 30 2 0 O-methylated (met) or non-methylated (nmet) piRNA. Equilibrium K<sup>D</sup> was calculated from steady-state binding levels Req = (CRmax)/(K<sup>D</sup> + C) + offset, where C = concentration, Rmax = analyte-binding capacity of the surface in response units (RU), K<sup>D</sup> = dissociation constant, and offset = response at zero analyte concentration. \* = *p* ≤ 0.05 by unpaired t-test with WT as comparison group.


In Hiwi1 PAZ, preferential binding of 30m RNA over 30nm RNA is mostly dictated by backbone confirmation of the protein rather than the amino acid composition of the binding pocket [44]. To investigate whether the mutations that impacted the RNA binding impacted the AePiwi4 PAZ secondary structure, we performed circular dichroism (CD) spectroscopy analysis with the T39A, Y40A, T41A, and T41R mutants and compared the CD curves to that of the WT AePiwi4 PAZ (Figure 4). We analyzed the data using CAPITO [51]. WT AePiwi4 PAZ displayed a CD curve most similar to proteins that had a mostly irregular structure but also that had between 30% and 49% beta strands and 6–16% alpha helices. All mutants maintained a mostly irregular secondary structure; however, they displayed different CD curves and percentages of alpha helices and beta-sheets compared to the WT protein (Figure 4 insets). T41R was most similar to proteins that were made up of between 16% and 26% beta strands and 28–46% alpha helices. Y40A also displayed a spectrum that aligned more with proteins that had a greater abundance of alpha helices—26–40% alpha helices but only 14–22% beta-strands. The T39A CD curve clustered with proteins that were made up of 9–25% alpha helices and 30–41% beta-strands, while the T41A curve clustered with proteins that were made up of 31–50% alpha helices and 4–21% beta-strands. Taken together, these data indicate that single amino acid changes in the AePiwi4 PAZ backbone can alter secondary structure, which likely impacted RNA-binding behaviors.

## *2.4. AePiwi4 Co-Localizes in the Cytoplasm and Nucleus in A. aegypti Tissues*

For further insights into the function of *A. aegypti* Piwi4, we characterized the subcellular localization of the native protein in *A. aegypti* mosquitoes. While this manuscript was under preparation, Joosten and colleagues (2021) reported that Piwi4, Piwi5, and Piwi6 were in both the nucleus and the cytoplasm in an *A. aegypti* embryonic cell line infected or uninfected with Sindbis virus [26]. To determine the sub-cellular localization of native AePiwi4 in both somatic and germline tissues in the mosquito, we generated polyclonal antibodies against AePiwi4 by immunizing mice with the AePiwi4 PAZ recombinant protein. We confirmed that these antibodies recognized both recombinant AePiwi4 PAZ and full-length proteins by Western blot (Supplemental Figure S5). To confirm that the antibodies recognized AePiwi4 from mosquito tissues, we prepared whole mosquito lysates from three *A. aegypti* females 48 h post-bloodmeal (time of peak AePiwi4 expression [32]) for Western blot. Anti-AePiwi4 mouse serum reacted to whole mosquito lysate at the expected size of AePiwi4 (100 kDa) (Supplemental Figure S5). Mass spectrometry analyses further confirmed that AePiwi4-specific peptides were present at the same location on a corresponding SDS/PAGE gel slice (Supplemental Dataset 1).

**Figure 4.** Circular dichroism (CD) spectra analyses of AePiwi4 PAZ mutants. CD spectra curves, by mean residue ellipticity for AePiwi4 PAZ WT and mutant proteins, recorded over 200–260 nm. Insets show the calculated percentages of secondary structures determined by CD analysis using CAP-ITO. Red bars indicate mean and standard deviation of three similarity hits based on lowest area differences under the curve. **Figure 4.** Circular dichroism (CD) spectra analyses of AePiwi4 PAZ mutants. CD spectra curves, by mean residue ellipticity for AePiwi4 PAZ WT and mutant proteins, recorded over 200–260 nm. Insets show the calculated percentages of secondary structures determined by CD analysis using CAPITO. Red bars indicate mean and standard deviation of three similarity hits based on lowest area differences under the curve.

*2.4. AePiwi4 Co-Localizes in the Cytoplasm and Nucleus in A. aegypti Tissues*  For further insights into the function of *A. aegypti* Piwi4, we characterized the subcellular localization of the native protein in *A. aegypti* mosquitoes. While this manuscript was under preparation, Joosten and colleagues (2021) reported that Piwi4, Piwi5, and Piwi6 were in both the nucleus and the cytoplasm in an *A. aegypti* embryonic cell line infected or uninfected with Sindbis virus [26]. To determine the sub-cellular localization of native AePiwi4 in both somatic and germline tissues in the mosquito, we generated polyclonal antibodies against AePiwi4 by immunizing mice with the AePiwi4 PAZ recombinant protein. We confirmed that these antibodies recognized both recombinant To determine the sub-cellular localization of AePiwi4, we next performed immunofluorescence assays and Western blots using both somatic and germline-derived mosquito tissues. AePiwi4 tended to stain cytoplasmically in both midguts (Figure 5A; Supplemental Figure S6A) and unfertilized embryos from ovary tissues (Figure 5B, Supplemental Figure S6B). However, when we fractionated ovaries from *A. aegypti* mosquitoes 48 h post-bloodmeal into cytoplasmic and nuclear fractions for Western blot, we found that AePiwi4 was present in both fractions (Figure 5C). Antibodies targeting H3 histone were used as a marker for the nuclear fraction (Supplemental Figure S7). These results suggested AePiwi4 may be trafficked in and out of the nucleus in mosquito tissues.

AePiwi4 PAZ and full-length proteins by Western blot (Supplemental Figure S5). To confirm that the antibodies recognized AePiwi4 from mosquito tissues, we prepared whole mosquito lysates from three *A. aegypti* females 48 h post-bloodmeal (time of peak AePiwi4 expression [32]) for Western blot. Anti-AePiwi4 mouse serum reacted to whole mosquito lysate at the expected size of AePiwi4 (100 kDa) (Supplemental Figure S5). Mass spectrometry analyses further confirmed that AePiwi4-specific peptides were present at the

To determine the sub-cellular localization of AePiwi4, we next performed immunofluorescence assays and Western blots using both somatic and germline-derived mosquito tissues. AePiwi4 tended to stain cytoplasmically in both midguts (Figure 5A; Supplemental Figure S6A) and unfertilized embryos from ovary tissues (Figure 5B, Supplemental

same location on a corresponding SDS/PAGE gel slice (Supplemental Dataset 1).

Figure S6B). However, when we fractionated ovaries from *A. aegypti* mosquitoes 48 h postbloodmeal into cytoplasmic and nuclear fractions for Western blot, we found that AePiwi4 was present in both fractions (Figure 5C). Antibodies targeting H3 histone were used as a marker for the nuclear fraction (Supplemental Figure S7). These results suggested

AePiwi4 may be trafficked in and out of the nucleus in mosquito tissues.

**Figure 5.** AePiwi4 expression in mosquito tissues. IFA of *A. aegypti* midgut (**A**) or ovaries with unfertilized embryos (**B**) stained with anti-AePiwi4 (red), phalloidin (green), or DAPI (blue). Scale bars are 30 μM and 20 μM for midguts and embryos, respectively; (**C**) Coomassie-stained SDS/PAGE gel (left) and corresponding Western blot of cytoplasmic (lane 1) or nuclear (lane 2) fractions of *A. aegypti* mosquito ovary tissue. **Figure 5.** AePiwi4 expression in mosquito tissues. IFA of *A. aegypti* midgut (**A**) or ovaries with unfertilized embryos (**B**) stained with anti-AePiwi4 (red), phalloidin (green), or DAPI (blue). Scale bars are 30 µM and 20 µM for midguts and embryos, respectively; (**C**) Coomassie-stained SDS/PAGE gel (left) and corresponding Western blot of cytoplasmic (lane 1) or nuclear (lane 2) fractions of *A. aegypti* mosquito ovary tissue.
