FACS-based isolation and RNA extraction of Secondary Cells from the Drosophila male Accessory Gland

To appreciate the function of an organ, it is often critical to understand the role of rare cell populations. Unfortunately, this rarity often makes it difficult to obtain material for study. This is the case for the Drosophila male accessory gland, the functional homolog of mammalian prostate and seminal vesicle. In Drosophila, this gland is made up of two morphologically distinct cell types: the polygonally-shaped main cells, which compose 96% of the organ, and the larger, vacuole-containing secondary cells (SCs), which represent the remaining 4% of cells (~40 cells per lobe). Both cell types are known to produce accessory gland proteins (Acps), which are important components of the seminal fluid and are responsible for triggering multiple physiological and behavioral processes in females, collectively called the post-mating response (PMR). While a few genes are known to be specific to the SCs, the relative rarity of SCs has hindered the study of their whole transcriptome. Here, a method allowing for the isolation of SCs is presented, enabling the extraction and sequencing of RNAs from this rare cell population. The protocol consists of dissection, protease digestion and mechanical dissociation of the glands to obtain individual cells. Then, the cells are sorted by FACS, and living GFP-expressing SC singulets are isolated for RNA extraction. This procedure is able to provide SC-specific RNAs from ~40 males per condition in the course of one day. Given the speed and low number of flies required, this method enables the use of downstream RT-qPCR and/or RNA sequencing to the study gene expression in the SCs from different genetic backgrounds, ages, mating statuses or environmental conditions. SUMMARY Here, we describe the dissociation and sorting of a specific cell population from the Drosophila male accessory glands (Secondary cells), followed by RNA extraction for sequencing and RT-qPCR. The dissociation consists of dissection, proteases digestion and mechanical dispersion, followed by FACS purification of GFP-expressing cells.

Accessory glands (AGs) are key components of the male reproductive tract in insects, being responsible for the production of most of the seminal fluid proteins (SFPs) that are known to induce the physiological and behavioral processes in females collectively called the post-mating response (PMR). The PMR includes, but is not restricted to, increased ovulation and egg-laying, sperm storage and release, diet changes and gut growth, and decreased receptivity to secondary courting males 1,2 . As such, AGs and SFPs are topics of intense interest to better understand the basic biological questions related to mating, reproduction, and evolution. Also, they have important impacts on major societal issues related to human health (some insects are vectors of deadly diseases) and agriculture (insects are both considered pests and critical for pollination and soil quality). Drosophila melanogaster is a prominent model for the study of AGs and ACPs, having brought many insights into the biology of these organs and the role of individual proteins regarding the PMR. The discoveries in fruit flies have largely affected the work in other species such as the disease vector Aedes aegypti 3,4 , and other insects 1,5 . Furthermore, the fact that the AGs secrete SFPs to be transferred to females during mating 1,6 makes the AGs the functional analog of mammalian prostate gland and seminal vesicle. Due to the functional and molecular similarities between the two tissue-tpyes, the AGs have been used as a model for the prostate gland in flies 7 .
Drosophila accessory glands are composed of two lobes consisting of a monolayer of secretory cells surrounding a central lumen, and wrapped by smooth muscles. The secretory cells comprise two morphologically, developmentally and functionally distinct cell types: most of the gland is composed of polygonally-shaped main cells (~96% of the cells), while larger and rounded secondary cells (SC), make up the remaining 4% (~40 cells per lobe). Both cell types produce distinct sets of ACPs and work interdependently to induce and maintain the PMR.
The major trigger of the PMR is the Sex Peptide, a small 36 amino acid protein secreted by main cells and known to cause most PMRs in females [8][9][10] . But many other ACPs produced by main and Secondary cells also affect different aspects of the PMR [11][12][13][14][15][16][17] . Based on our current knowledge, it seems that SCs and their products are required to perpetuate the effects of SP past one day 18 .
Thus far, most of the knowledge that we have accumulated on SC biology comes from candidate approaches, finding the expression of one particular gene or protein in these cells and determining its role in the development and/or function of the SCs. These genes include the homeodomain protein Defective proventriculus (Dve, 19 ), the lncRNA MSA 20 , Rab6, 7, 11 and 19 21 , CG1656 and CG17575 11,15,21 and the homeobox transcription factor Abdominal-B (Abd-B) 18 . A mutant line deficient for both the expression of Abd-B and the lncRNA MSA in secondary cells has been used to determine that secondary cells are required for Sex Peptide to be properly stored in the female reproductive tract, which results in a shortened PMR (from ~10 days to one day) 12,18,20 . At the cellular level, the SC of this mutant almost lose their characteristic vacuole like structures 18,20,21 . This mutant line was successfully used to identify some genes involved in these phenotypes by comparing transcriptomes of wild type versus mutant accessory glands 12 .
Unfortunately, it was difficult to access the full genetic program of SC, because of their relative rareness in an organ made up primarily of main cells. For qPCR validation of genes suspected to be induced under particular conditions, the abundance of main cells would often hide the variation in gene expression, and performing in situ hybridization on glands proved to be tricky. We thus decided to develop a method for isolating purified SC RNA that was easy and fast enough to perform in a variety of different genetic backgrounds or environmental conditions.
Abd-B and MSA expression in SCs relies on the D1 enhancer, a 1.1kb piece of DNA located in the iab-6 regulatory region of the Bithorax complex 18,20 . GAL4 drivers containing this sequence are expressed in secondary cells and, when associated with a UAS-GFP, give a strong GFP signal in live SCs, allowing clear visualization and FACS sorting of these cells. The iab-6 cocuD1 chromosome has a small deletion of this specific D1 enhancer, abrogating Abd-B and MSA expression in SC, and causing the phenotypes described above 20 . We performed this protocol on wt and iab-6 cocuD1 mutant accessory glands as a proof of principle that this approach can not only provide the wild type transcriptome of this rare cell type, but also to identify mis-regulated genes involved in SC function.

1.
Drosophila line generation and male collections

REPRESENTATIVE RESULTS:
The method presented here allows, in the course of one day, for isolating GFP-expressing secondary cells from Drosophila accessory glands, and extracting their RNA for sequencing. Figure 1: Overview of the protocol Key steps of the protocol are shown, with the timeline on the right side. This procedure allows one starting with live Drosophila in the morning to have dissociated accessory gland cells by noon, sort them based on GFP expression, and get their RNAs extracted by the end of the working day. RNA sequencing and data analysis will typically take a few weeks.
We use here the Abd-B-GAL4 construct described in 18 to express GFP in secondary cells (SC) but not in main cells (MC) (Figure 2A). The first objective of the method is to get the transcriptome of "wild type" SC (wild type is put in quotation marks because these flies are transgenic animals carrying a GAL4 driver express GFP). The second objective is to be able to obtain SC RNA quickly and easily enough so it is possible to compare their transcriptomes in different conditions. To test for this, we performed this protocol from "wild type" and "iab-6 cocuD1 " mutants carrying a deletion of 1.1kb removing the SC enhancer of Abd-B as well as the promoter of the MSA transcript, known to be critical for SC development, morphology and function 18,20 (Figure 2B and C). We repeated this protocol 3 times on different days to generate the triplicates of each genotype presented here throughout the figures (hereafter referred as wt-1,-2,-3 for the wild type and D1-1,-2 and -3 for the iab-6 cocuD1 ).
The method described herein allows dissociating MC and SC from Drosophila accessory glands in only a few hours ( Figure 2D). These cells are then sorted by FACS into two distinct tubes to isolate MCs and SCs. The FACS gating strategy is presented in Figure 2E to G. The addition of Draq7 allows the estimation cell viability, which is around 70% for the whole sample ( Figure 2F). The singulets -estimated by double gating on FSC-A vs FSC-H and SSC-H vs SSC-W -represent over 90% of the SC population, and over 80% of the MC population, reflecting the efficiency of dissociation (exemplified in Figure 2H). 10 to 12% of SC sortings were aborted due to the presence of another cell or debris in the droplet. Starting with 40 males, we typically stop sorting after collecting 550 SCs and 1000 MCs. With one out of 6 samples, we did not reach the objective of 550 SC: the wt-1 sample was prepared with only 427 SC, but gave equally good results.
SC and MC are sorted into lysis buffer and RNA extraction is performed as soon as all samples are ready in order to obtain RNA pellets by the end of the same day. RNA quality and quantity are estimated on a Bioanalyser using an appropriate chip to work with small volumes and low concentrations. Since estimated concentrations were quite variable between samples (ranging from 344pg/μl to over 1300pg/μl, see Figure 3A), the starting material for both RT-qPCR and cDNA library synthesis was roughly adjusted to 2ng (the measured concentration is not very precise). The expression of specific genes was quantified by Real Time qPCR on SC and MC extracts of the "wt" condition to control for the identity of the sorted populations. The results shown in Figure 3B are relative quantifications normalized to alpha-tubulin for each sample. As expected, housekeeping genes alpha-tubulin and 18S rRNA are detected in all samples. On the contrary, the SC-specific gene rab19 is only detected in SC extracts, while the MC-specific gene Sex Peptide is only detected from MC. As expected, the SC-specific transcript MSA whose promoter is deleted in Iab-6 cocuD1 is detected only from wt SC and not from mutant SCs.
Together, the quality controls presented in Figure 3 show that the RNAs obtained with this procedure are not degraded and that SC and MC populations are successfully sorted, from both wt and mutant accessory glands. We sequenced only secondary cells' RNAs, but note that this protocol allows for concomitant RNA profiling of both SC and MC.
The RNA-sequencing was performed using standard protocols. For the purpose of this method, we will only discuss here the pertinent quality control analysis. After sequences were obtained, the reads were mapped to the reference genome, attributed to genes, counted and normalized. Principal Component Analysis (PCA) was performed on all 6 samples (three "wild type" replicates and three iab-6 cocuD1 replicates). PC1 accounts for as much of the variability in the data as possible, and PC2 accounting for as much of the remaining variability as possible. As presented in Figure 4A, the 3 wt replicates cluster together, and far away from Iab-6 cocuD1 samples, showing that wt samples are similar to each other, but different from mutant samples. This shows the reproducibility of the method, and its ability to characterize the divergent genetic program of mutant SC. We note that while D1-2 and D1-3 samples cluster together, the D1-1 sample is quite different. As all quality controls for this sample are good and similar to all other samples, we can exclude a problem in sample preparation (29 million total reads of which >76% align uniquely to the genome and >77% of them are attributed to a gene, >90% mRNAs, <3% rRNA). This variance could thus reflect that gene expression in iab-6 cocuD1 SC is unstable, although more replicates would be necessary to test this hypothesis.
Visualization of reads on the genome at particular genes allows a direct and visual estimation of the quality of the data. In Figure 4 a selection of genes is shown, with one representative replicate of each genotype. As expected housekeeping genes such as Act5C ( Figure 4B) are expressed in both conditions, as well as the SC genes Rab19 and Dve ( Figure 4C-D). The absence of reads in introns shows that polydT reverse transcription successfully selected the mature spliced mRNAs to prepare cDNA library. Importantly, we can see strong and significant variations in some genes' expression levels comparing wild type and iab-6 cocuD1 . This is exemplified by the MSA gene presented in Figure 4E whose expression is strong in wild type and absent from iab-6 cocuD1 . This gene is shown as a proof of principle that this method allows identifying significantly mis-regulated genes which could shed light on the mechanisms responsible for the phenotypes observed in this mutant and give new insights into normal SC function.   Panels B to E show sequencing reads mapped to the Drosophila reference genome using the IGV software. Only one representative sample of each genotype is shown for clarity sake (wt-1 and D1-2), and only a few specific loci are shown. Gene names are written on top of each panel, > and < symbols refer to their orientation. Numbers in brackets represent for each track the scale for the number of reads per DNA base pair. This scale is the same for both conditions for a given gene, but varies between genes for better visualization. Blue bars at the bottom of each panel show genes' introns (thin line), exons (wide line) and ORF (rectangles with >>). Note that Rab19 and Arl5 are overlapping, convergent genes (C). Secondary cells represent a minor cell type of the Drosophila accessory gland, yet play a critical role in male fertility by maintaining the post-mating response in females. In this protocol, we present a quick and efficient method to access the full transcriptome of these cells that account for only 4% of the organ, i.e. ~80 cells per individual. This method is based on dissection, peptidase digestion and FACS sorting, and can be performed in one day for multiple samples (except the RNA sequencing part).

Table1: Primers sequence
Methods to dissociate cells from Drosophila tissue such as imaginal discs were described previously 23 . However, our attempts to simply use these methods with accessory glands failed, prompting us to develop this method. For a successful dissociation, the peptidases have to access the accessory gland cells, which are protected by an outer muscle layer and an inner viscous seminal fluid. Thus, cutting the glands open (step 3.6.) and digesting for at least 60 minutes with vigorous shaking (step 4.1.) is critical to success. TrypLE allowed gentle dissociation, conserving the integrity of the gland and viability of the cells until mechanical dissociation. While papain and collagenase were not sufficient to dissociate the cells, we found that both of those enzymes improved the dissociation in association with TrypLE (less pipetting was required to obtain perfect dissociation, resulting in better survival). The trituration with narrow, rounded tips (steps 5.2. and 5.3.) is a key step and should be optimized in pilot experiments due to the way these tips are created (see step 2.3.2. and the Note after step 5.3. for guidelines).
This method allows one to isolate around 500-800 individual, live Secondary cells from 40 males (we stop sorting around 550 cells per sample to normalize material for RNA extraction). This corresponds to ~20% (±5%) efficiency assuming a starting material containing ~ 3200 SC (40 males x 80 SC). 20% recovery was enough for our purpose as we could obtain several samples in one day. However it might be improved by different means, including: working in larger batches and reducing transfers (doing trituration in the digestion tube, skipping step 5.4. and going straight to the FACS); performing a more gentle trituration (a significant proportion of dissociated GFP+ cells visible after step 5.3. die in the hour following trituration, as they are probably damaged in the process); reducing the digestion duration (using a more concentrated TrypLE might be an attractive option); reducing the stringency of singulet selection (step 6.2.3.); reducing time between time between dissociation and FACS… Secondary cells have unique cell morphology, having two polyploid nuclei and a vacuole-filled cytoplasm, consistent with a role in producing, modifying, and secreting a large quantity of proteins into the seminal fluid 24 . These secretory cells fulfill non-redundant functions essential for male fecundity 12,20,25 , and having a global view of all the genes that they express gives new insights into their normal function. We show here that the transcriptome of these cells can be obtained from a relatively small number of males, allowing for the comparison of different conditions. Here we use one mutant known to affect secondary cells morphology and function, and show that its transcriptome is significantly changed, suggesting that this method will allow identification of new important secondary cell genes. Previously in the lab a similar analysis of wt and iab-6 cocuD1 SC had been done by manually picking the cells, and produced good quality RNA sequencing data as well, but the method was too labor intensive to be extended to other conditions. In a separate manuscript, currently in preparation, we will extensively present and compare our datasets, and more importantly analyze them regarding their biological significance and how they help us understanding the normal function of SC.
As the morphology, vacuolar content and the number of Secondary cells has been shown to change with age, mating status, and diet 21,26,27 , it would be interesting to compare SCs under different conditions. Having a simple and fast protocol will allow one to study SC under all of these different conditions. It is noteworthy that this protocol allows simultaneous isolation of main cells from the same individuals, and thus could also be used without modification to see how the different genetic and environmental parameters affect mains cells at the same time.