Efficient Expression of Genes in the Drosophila Germline Using a UAS Promoter Free of Interference by Hsp70 piRNAs

Using the yeast GAL4 transcription factor to control expression in Drosophila melanogaster has long been ineffective in female germ cells during oogenesis. Here, DeLuca and Spradling show that the expression problem of most Drosophila molecular tools...

D ROSOPHILA is an extremely powerful model organism for studies of animal development and disease because of its low maintenance costs, rapid generation time, and expansive collection of tools to genetically modify its cells. One particularly useful tool is the Gal4/upstream activation sequence (UAS) two-component activation system, in which the Gal4 transcriptional activator protein recognizes a UAS to induce the expression of any gene of interest (Fischer et al. 1988;Brand and Perrimon 1993). By controlling the activity of Gal4 with tissue-specific or inducible promoters, or the Gal80 inhibitor protein, one can manipulate genes in specific cells or times of development, visualize cell types, probe cell function, or follow cell lineages. One of the most useful applications of these techniques has been to carry out genetic screens by expressing RNA interference (RNAi) in targeted tissues or cultured cells (Dietzl et al. 2007;Ni et al. 2008).
The original pUASt vector from Brand and Perrimon (1993), which contains an Hsp70-derived core promoter and simian virus 40 terminator, has undergone several optimizations to improve its expression ( Figure 1A). Popular versions, such as the Valium10 or 20 vector used by the Drosophila Transgenic RNAi project (TRiP) (Ni et al. 2009(Ni et al. , 2011 and the pMF3 vector used by the Vienna Drosophila Research Center (VDRC) GD collection (Dietzl et al. 2007) added a ftz intron, and the Janelia Gal4 enhancer project used derivatives of pJFRC81, which added a myosin IV intron (IVS), synthetic 59-UTR sequence (syn21), and viral p10 terminator to boost expression levels across all Drosophila cell types ( Figure 1A) (Pfeiffer et al. 2012). However, these modifications did not correct UASt's major problem, that it drives woefully poor expression in the female germline compared to somatic tissues. Consequently, genetic manipulation in this important tissue has often relied on a special GAL4-activated promoter, UASp, produced by fusing 17 copies of the UAS activator to a germline-compatible promoter derived from the P-element, a transposon naturally active in the female germline ( Figure 1B) (Rørth 1998). Although UASp expression is qualitatively higher than UASt in the female germline, it is generally known to be lower in somatic tissues.
The lack of a UAS construct that is widely useful in all Drosophila tissues has remained an obstacle to providing optimum genetic tools to the research community. Transgenic RNAi collections were first constructed using UASt, and screening of genes for germline functions has relied on increasing the effectiveness of RNAi by coexpressing Dcr2 or expressing short hairpin RNAi from UASp promoters (Ni et al. 2011;Yan et al. 2014;Sanchez et al. 2016). A significant obstacle to obtaining a widely effective GAL4 vector has been the lack of understanding of the reason that UASt functions poorly in germ cells, and the paucity of accurate comparisons between the UASp and UASt promoters in the absence of other significant variables.

New stocks created for this study
Bestgene introduced pMRtGFP and pMRpGFP into yw flies using a P-transposase helper plasmid, and we isolated GFP+ insertions by crossing the F0 to a Mef2-Gal4 background and scoring for GFP+ muscles. We introduced UAStGFP or UASpGFP into MI04106 and other MiMIC lines using a cross strategy outlined in Nagarkar-Jaiswal et al. (2015). Rainbow transgenics introduced pJFRC81 (UAStGFP-attB), pUASpGFP-attB, and pUASzGFP-attB into attP40 using an X chromosome-encoded fC31 integrase source, and we isolated multiple w + , fC31 minus insert lines by standard fly genetics.

Vectors created for this study
Genescript synthesized pMRtGFP. We created pMRpGFP by replacing the NheI-BglII UASt promoter in pMRtGFP with a SpeI-BglII UASp promoter from Valium22. We created pUASpGFP-attB by replacing the PstI-BglII UASt promoter in pJFRC81 with the PstI-BglII UASp promoter from Figure 1 Components of common upstream activator sequence (UAS) constructs used by the fly community. (A) Cartoon depicting a Drosophila Hsp70 gene relative to sequences in UASt-based vectors. In pUASt, multiple copies of optimized Gal4-binding sites (5xUAS) replace heat-inducible enhancers (Heat Shock Elements, HSEs) in a fragment of Hsp70 containing the transcription start site (TSS) and 59-UTR. In derivatives of UASt, a multiple cloning site (MCS), RNA interference (RNAi) constructs, GFP coding sequence, synthetic UTR elements (syn21), and introns (ftz or myosin IV, IVS) replace 39 bp of the Hsp70 59-UTR and Hsp70 coding sequence (CDS). Viral-derived simian virus 40 (SV40) or p10 sequences terminate transcription and contribute to the 39-UTR. For this study, we created a derivative of pJFRC81 with a truncated 59-UTR (pUASzGFP-attB) and a derivative compatible with MiMIC recombinase-mediated cassette exchange (pMRtGFP). (B) Cartoon depicting two common UASp vectors containing the K10 terminator and Drosophila P-element promoter, TSS, and 59-UTR in place of the SV40 terminator and Hsp70 sequences. We created two new UASp vectors, pUASpGFPattB and pMRpGFP, based on pJFRC81and pMRtGFP, to directly compare the effect of P-element and Hsp70 sequences on transgene expression. Vector names colored red are used in this study. TRiP, Drosophila Transgenic RNAi project; VDRC, Vienna Drosophila Research Center.
Valium22. We created UASzGFP-attB by replacing the 259-bp NheI-BglII fragment of pJFRC81 containing the 203-bp Hsp70 promoter with annealed oligos encoding 63 bp from the 59 end of the same promoter.
We created UASz1.1 by replacing the KpnI-EcoRI p10 terminator in UASz with a PCR-amplified p10 terminator containing Kpn1-XbaI-EcoRI and ApoI tails. We created UASz1.2 by destroying the NheI and EcoRI sites in UASz1.1 by cloning annealed oligos into the NheI-EcoRI backbone.
To create UASzMiR, we cloned a BglII-XhoI fragment containing the MiR1 cassette and ftz intron from Walium22 into the BglII-XhoI backbone of UASz1.2.

Tissue preparation, imaging, and quantitation
For all experiments, we crossed UAS-GFP or UAS-GFP Hsp70D males to control (yw), Tub-Gal4/TM3, homozygous Vasa-Gal4, or homozygous Vasa-Gal4 Hsp70D females. For whole-larvae imaging, we picked wandering third-instar larvae of various genotypes, aligned them on the same glass slide, and placed them the freezer for 30 min prior to imaging. For adult ovary or larval tissue imaging, we fixed dissected tissue with 4% paraformaldehyde for 13 min (whole ovary) or 20 min (larval tissue) and stained with DAPI in PBS with 0.1% Triton X-100. We imaged the GFP fluorescence of semifrozen whole third-instar larvae or whole ovaries mounted in 50% glycerol on a Leica Stereoscope equipped with a mercury arc light source, GFP filters, and a CCD camera. We imaged GFP fluorescence in larval imaginal discs, salivary glands, and epidermis, and manually separated ovarioles mounted in 50% glycerol using a custombuilt spinning disc confocal with a 203 0.8 NA lens. For each genotype and tissue type, we acquired a single-plane image from at least four individuals using Metamorph software and the same laser power, CCD camera gain, and exposure time between equivalent samples. We measured average pixel intensity in 14-bit images of the GFP channel using Image J. We acquired representative images of single planes through single ovarioles for Figure 2 on a Leica Sp8 scanning confocal with a 633 1.4 NA lens and PMT (for DAPI) and HiD (GFP) detectors using identical settings between samples.
For UASt Piwi-interacting RNAs (piRNA) analysis, we clipped and aligned sequenced small RNA libraries from Mohn et al. (2014) (SRR1187947:control germline knockdown and SRR1187948:rhino germline knockdown) to the Drosophila melanogaster genome Release 6 (Hoskins et al. 2015) or UAStGFP using the Bowtie2 aligner with no filtering for repetitive mappers (Langmead and Salzberg 2012). We visualized piRNA read depth to UAStGFP or both Hsp70 clusters using the Interactive Genome Browser (Robinson et al. 2011).

Data availability
Fly strains and vectors are available upon request. pUASz1.0 and pUASzMiR, and sequences, are available from the Drosophila Genomics Resource Center as items 1431 and 1432. Supplemental material available at Figshare: https://doi.org/ 10.25386/genetics.6089828.

Difference between UASp and UASt
To study the difference between the UASp and UASt promoters, we first created UAStGFP and UASpGFP constructs controlled for other variables between the original UASt and UASp, such as UTR components, introns, terminators, and genomic insertion site. Both constructs were based on pJFRC81 and only varied at the promoter and 59-UTR of the transcript (Figure 1, red letters). We made pMRtGFP and pMRpGFP compatible with fC31-catalyzed recombinationmediated cassette exchange with MiMIC transposons, allowing us to integrate UAS-GFPs into many common sites throughout the genome (Venken et al. 2011). Using a previously established protocol (Nagarkar-Jaiswal et al. 2015), we recombined both UAS-GFPs into several MiMICS, including MI04106, which resides in a region enriched for ubiquitously expressed genes and active chromatin marks (Filion et al. 2010;Kharchenko et al. 2011), referred to as "the gooseneck" by Calvin Bridges for its long stretch of low density in salivary gland polytene chromosome preparations (Bridges 1935). Consistent with previous reports, UASt drove significantly stronger expression than UASp in all somatic tissues examined, while UASp drove significantly stronger expression in the female germline (Figure 2, A and B).

Hsp70 piRNAs repress UASt
We next investigated the reason for the extremely weak UASt expression in the female germline. Several lines of evidence implicated piRNA-directed silencing as a mechanism limiting UASt expression. Drosophila piRNAs are ovary -and testisenriched, 23-29-nt RNAs that complex with Argonaut family proteins and silence transposons through homologous base pairing-directed mRNA cleavage and heterochromatin formation (Siomi et al. 2011). Some of the most successful UASt-based genetic screens in the female germline knocked down piRNA biogenesis genes (Czech et al. 2013;Handler et al. 2013). If piRNAs were silencing UASt, then UASt-RNAi against a piRNA biogenesis gene would boost UASt expression leading to maximal knockdown. Where might these UASt-piRNAs originate from? Previously, Olovnikov et al. (2013) characterized an abundance of germline-specific piRNAs mapping to both Hsp70 gene clusters. Because UASt contains the Hsp70 promoter and 59-UTR, we hypothesized that germline piRNAs against Hsp70 may be targeting UASt. When we searched for UASt sequences in the piRNAs identified by Mohn et al. (2015), we identified abundant piRNAs perfectly homologous to UASt ( Figure 2E, pink bars, and Figure 2F). Similar to UASt silencing, these UASt piRNAs are restricted to the female germline because germline-specific knockdown of rhino, a gene required for Hsp70 piRNA production, eliminates UASt piRNAs from whole ovaries ( Figure 2E) (Mohn et al. 2014).
To directly test whether Hsp70 piRNAs silence UASt, we tested UASt expression in Hsp70D flies (Gong and Golic 2004), which completely lack all genetic loci producing piRNAs homologous to UASt ( Figure 2E, gray boxes deleted). Despite missing all copies of the inducible Hsp70 gene family and related piRNAs, Hsp70D flies have no significant defects in viability or egg production in the absence of heat stress (Gong and Golic 2006). However, Hsp70D flies showed greatly enhanced UAStGFP expression. Furthermore, UAStGFP expression was significantly stronger than that of UASpGFP, which was unaffected by Hsp70D ( Figure 2D). Repression of full UAStGFP expression in germ cells requires normal piRNA production, since UAStGFP expression was also boosted by germline knockdown of rhino, which is required for the production of Hsp70 and many other germline-specific UASt-wRNAi aligned to our UAStGFP construct to show the mapping position of previously described piRNAs that could theoretically silence the transgenes used in this study. The green shaded area shows the 184 bp of UASt deleted to make UASz. piRNA, Piwi-interacting RNAs; RNAi, RNA interference; UAS, upstream activator sequence. piRNAs ( Figure 2C). These results argue strongly that UASt is normally silenced by Hsp70 piRNAs and that UASt is a stronger expression vector than UASp in cells lacking Hsp70 piRNAs.

Construction of UASz
We next attempted to create a new version of the UAS expression vector that works well in both the soma and the female germline. We hypothesized that eliminating the part of UASt that is targeted by piRNAs would boost UASt expression by the same amount as eliminating the piRNAs themselves. Hsp70 piRNAs are homologous to 247 nt of the UASt promoter and 59-UTR. While we could make enough substitutions along this stretch to prevent all possible 23-nt piRNAs from binding, we were afraid that this approach might impair important promoter sequences. Instead, we hypothesized that Hsp70 piRNAs might recognize UASt RNA to initiate piRNA silencing. To prevent Hsp70 piRNAs from recognizing UASt RNA, we trimmed down the UASt 59-UTR to be shorter than a single piRNA, from 213 to 19 nt ( Figure 1A and Figure  2E). We named this UTR-shortened UASt variant "UASz," because we optimistically hoped that it would be the last one anyone would make.

Comparison of UAS vectors
To compare the relative expression levels of our UASz to UASp and UASt, we created all three variants in the same GFP vector backbone (pJFRC81) with a single attB site. We used fC31 integrase to introduce these UAS-GFP variants into a commonly used genomic site, attP40, and recombined all three inserts with Hsp70D to determine the influence of Hsp70 piRNAs on their expression. When combined with Tub-Gal4, a somatic Gal4 driver, UASz was expressed at least four times higher than UASp in all somatic tissues tested and was equivalent or greater than UASt in some somatic tissues, like the larval epidermis and salivary gland (Figure 3, A, C, and E). However, UASz was expressed at 40% of UASt in discs, suggesting that some elements of the UASt 59-UTR may boost expression in some tissues (Figure 3, C and E). To measure germline expression, we crossed the three UAS-GFPs to vasa-Gal4, which is evenly expressed up to stage six of oogenesis. In the germline, UASz was expressed around four times higher than UASp at all stages, while UASt was expressed at much lower levels than UASp, except in region 1 of the germarium (Figure 3, B and D-F), where piRNA silencing is weaker (Dufourt et al. 2014). We conclude that UASz is a superior expression vector to UASp in all tissues, and is equivalent to UASt in many, but not all, somatic tissues.
Finally, we wanted to test if UASz is still targeted by Hsp70 piRNAs because it contains 63 nt of Hsp70 sequence and 10% of the putative piRNAs targeting UASt ( Figure 2F). We crossed UASzGFP into the Hsp70D background and compared UASzGFP levels with or without Hsp70 piRNAs. We observed no enhancement of UASzGFP when Hsp70 piRNAs were removed (Figure 3, B, D, and F). Therefore, Hsp70 piRNAs likely target the UASt but not the UASz 59-UTR, consistent with the model that piRNAs must initially recognize RNA but not DNA.
Is UASz the final, fully optimized iteration of a UAS vector? Probably not. UASt without Hsp70 piRNAs induces about twice the expression of UASz in the ovary (Figure 3, B, D, and F). This twofold advantage of UASt over UASz in the germline or imaginal discs lacking Hsp70 piRNAs is similar to the twofold advantage of UASt over the UAS fused to the Drosophila Synthetic Core Promoter (Pfeiffer et al. 2010). Perhaps adding back some sequences within the first 203 nt of the Hsp70 59-UTR while avoiding piRNA recognition may improve UASz. However, the current iteration of UASz remains an unequivocal upgrade over UASp for all applications and UASz should be preferred over UASt if both germline and soma studies are planned from a single vector. Alternatively, one could boost germline expression of an existing UASt construct by crossing it into the Hsp70D background.
Current UAS-RNAi collections are heavily biased toward UASt-RNAi-based constructs. To date, the VDRC and DRSC/ TRiP RNAi projects have used UASt-RNAi to target 12,539 and 8876 genes, respectively. Germline screens for developmental phenotypes using UASt-RNAi were enriched for phenotypes in germline region 1 (Yan et al. 2014;Sanchez et al. 2016), where piRNA silencing is weakest (Dufourt et al. 2014), and UASt shows maximum expression ( Figure 3D arrow). Perhaps these screens were depleted for developmental defects in later germline stages because of poor UAS-RNAi expression in these stages. Although UASp-RNAi from the Valium22 vector ( Figure  1B) increased the efficiency of obtaining phenotypes in a germline screen, only 1596 genes are currently targeted by this collection (Yan et al. 2014). Additionally, when screening somatic cells, Ni et al. (2011) recommend UASt-RNAi because UASp-RNAi gave incomplete knockdowns. Our results revealed that UASp is equally weak in the germline as somatic tissues when compared to UASz ( Figure 3E). Therefore, UASp-RNAi may also generate incomplete knockdowns in the germline. To increase germline RNAi expression, we propose using a UASzbased RNAi expression vector, such as UASzMiR (Supplemental Material, Figure S1), which is compatible with previously generated short hairpin RNA oligo cloning (Ni et al. 2011).