Increasing cell density globally enhances the biogenesis of Piwi-interacting RNAs in Bombyx mori germ cells

Piwi proteins and their bound Piwi-interacting RNAs (piRNAs) are predominantly expressed in the germline and play crucial roles in germline development by silencing transposons and other targets. Bombyx mori BmN4 cells are culturable germ cells that equip the piRNA pathway. Because of the scarcity of piRNA-expressing culturable cells, BmN4 cells are being utilized for the analyses of piRNA biogenesis. We here report that the piRNA biogenesis in BmN4 cells is regulated by cell density. As cell density increased, the abundance of Piwi proteins and piRNA biogenesis factors was commonly upregulated, resulting in an increased number of perinuclear nuage-like granules where Piwi proteins localize. Along with these phenomena, the abundance of mature piRNAs also globally increased, whereas levels of long piRNA precursor and transposons decreased, suggesting that increasing cell density promotes piRNA biogenesis pathway and that the resultant accumulation of mature piRNAs is functionally significant for transposon silencing. Our study reveals a previously uncharacterized link between cell density and piRNA biogenesis, designates cell density as a critical variable in piRNA studies using BmN4 cell system, and suggests the alteration of cell density as a useful tool to monitor piRNA biogenesis and function.


Results
The abundance of Piwi proteins and piRNA biogenesis factors is linked to cell density. During the course of investigating the piRNA pathway using BmN4 cells, we noticed that the levels of Piwi proteins and piRNAs differed in cells with varying densities. To clarify the relationship between cell density and the piRNA pathway, we plated BmN4 cells with varying cell densities, cultured them for 30 h, and then harvested the cells for the analyses of protein and RNA expression ( Fig. 1A and B). Western blots, with quantitative ability (Supplementary Fig. S1 and S2), showed that the abundance of both of the Piwi proteins, Siwi and BmAgo3, as well as the factors involved in piRNA biogenesis, BmVasa and BmPapi, commonly increased along with increasing cell density (Fig. 1C). No obvious quantitative changes were observed in the control cytoplasmic (β-actin) and mitochondrial (Hsp60 and Tom20) proteins (Fig. 1C), in total protein staining patterns ( Supplementary Fig. S3), and in mitochondrial staining patterns ( Supplementary Fig. S4), suggesting that the increases in expression levels may be specific to the protein factors involved in piRNA biogenesis.
To examine whether the increased accumulation of piRNA biogenesis-related factors is caused by transcriptional upregulation of their mRNAs, we quantified their mRNA expression levels by qRT-PCR. As shown in Fig. 1D, none of their mRNA levels was affected by cell density, implying that the increased accumulation of the piRNA biogenesis proteins with increasing cell density occurs through post-transcriptional mechanisms, such as translational activation or increased protein stability.
The observed increase in accumulation of piRNA biogenesis-related factors could be triggered by one of many potential events occurring in the cells with increased cell density. The different expression levels of Siwi in highand low-density cells were retained when the cells with the two densities were cultured in the same medium using the Transwell system (Fig. 1E), indicating that the increased accumulation in high-density cells is not due to the confounding effects of nutrients or other diffusible factors in the cultured medium. In addition, although the increase of cell density accelerates the rate of cell proliferation ( Supplementary Fig. S5A), the Piwi protein levels are not affected in the cells in which the cell cycle was arrested by double thymidine block (Supplementary Fig. S5B and C), implying that the increased cell proliferation rate is not a significant factor for the increased accumulation of piRNA biogenesis factors in high-density cells. These results suggest that physical cell-cell contact, but not cultured medium or cell proliferation, contributes to the upregulation of the abundance of piRNA biogenesis factors.
Cell density-dependent formation of perinuclear nuage-like granules. Piwi proteins accumulate in amorphous, ribonucleoprotein-rich, perinuclear granules that are named nuage in Drosophila and intermitochondrial cement or chromatoid body in mice 4,45 . In BmN4 cells, Siwi, BmAgo3, and BmVasa are mainly co-localized in perinuclear nuage-like granules 33,[36][37][38] (Supplementary Fig. S1B and C). To examine the influence of cell-cell contact on the granule formations, the three proteins were immuno-stained in cells of varying densities. As shown in Fig. 2A and B, the number of perinuclear granules and protein signal intensities drastically increased along with increasing cell density, suggesting that increasing cell density promotes granule formation and accumulation of the piRNA biogenesis factors in the granules.
Global increase in mature piRNA abundance with increasing cell density. We subsequently investigated the expression levels of mature piRNAs in BmN4 cells with varying densities. Northern blot for the piR-1 and piR-2, which specifically bind to Siwi and BmAgo3, respectively 36 , showed that the expression levels of both piRNAs were commonly enhanced by increasing cell density ( Fig. 3A and B). The results were confirmed by piRNA quantification using qRT-PCR with a stem-loop primer ( Supplementary Fig. S6), which was modified from stem-loop qRT-PCR for miRNA quantification 46 . As observed in Northern blot, qRT-PCR quantification revealed increased accumulation of both piR-1 and piR-2 in high-density cells (Fig. 3C), suggesting that the abundances of both the Siwi-and BmAgo3-bound piRNAs, as well as the Siwi and BmAgo3 proteins, are linked to cell density. Because increasing cell density has been shown to globally up-regulate the expression of miRNAs in human cancer cells 44 , we also quantified the levels of miRNAs. All 6 examined miRNAs commonly showed enhanced accumulation in high-density cells ( Supplementary Fig. S7), suggesting that not only piRNAs but also miRNAs are under control of cell density.
Scientific RepoRts | 7: 4110 | DOI:10.1038/s41598-017-04429-7 To investigate whether piRNA precursors are also regulated by cell density, we quantified a reliable piRNA precursor expressed from Torimochi, a representative functional piRNA cluster in the Bombyx genome 35 . The Torimochi piRNA precursor contains a 5′-Cap and 3′-Poly(A)-tail, and produces abundant piRNAs in BmN4 cells 35 . In contrast to those of piRNAs, the expression levels of the Torimochi precursor decreased as cell density increased (Fig. 3D). We reason that the decreased levels could result from global enhancement of piRNA biogenesis mechanisms in higher density cells and increased use of precursors for piRNA production.
To globally assess the influence of cell density on piRNA populations, we performed SOLiD next-generation sequencing (NGS) of small RNA species of BmN4 cells with low, medium, or high densities. In staining patterns of total RNA extracted from the respective cells, we confirmed the increasing abundance of ~27-29-nt mature piRNA populations in higher density samples, whereas other RNAs (e.g., tRNAs) showed no apparent quantitative differences between samples (Fig. 4A). NGS yielded approximately 118, 145, and 152 million total reads from the low, medium, and high samples, respectively. After mapping to the B. mori genome, majority of the 16-50-nt reads (87.5%, 90.1%, and 92.8% of the reads from the low, medium, and high samples, respectively) were derived from the regions of miRNAs (19-23 nt) and piRNAs (24-30 nt). Given that not only piRNAs but also miRNAs showed increased levels in high-density cells, quantitative comparison of the reads between the samples was not possible due to a lack of a normalization method. Consequently, we utilized the NGS data for qualitative comparison of the reads and for comparison of piRNA populations. The relative proportions of known annotations Ribosomal protein 49 (Rp49) was used as an internal control 36 . mRNA levels in the cells with 6.0 × 10 3 cells/ cm 2 starting density were set as 1. Each data set represents the average of three independent experiments with bars showing the SD. (E) The low-and high-density cells, which were co-cultured in the same medium using a transwell system, were subjected to Western blots with indicated antibodies. of the reads were not appreciably varied between samples (Fig. 4B). Nucleotide composition analyses showed strong piRNA characteristic biases for uridine on the 5′-end (position 1) and adenine on the position 10, but no difference was observed between samples (Fig. 4C). Moreover, in the low, medium, and high samples, we detected similar patterns of ping-pong signals (sense-antisense piRNA pairs overlapping by 10 nt at their 5′-ends that are characteristic of a ping-pong amplification cycle in secondary piRNA biogenesis 4 ) (Fig. 4D). Mapping patterns of the reads to transposon sequences were also similar between samples ( Supplementary Fig. S8). To confirm the global influence of cell density on piRNA abundance, we selected 16 mature piRNAs (Supplementary Table S2) from the reads meeting the following criteria: (1) 27-29 nt in length, (2) among the top 100 abundant reads in at least two of the three libraries (low, medium, and high), and (3) among Siwi-or BmAgo3-bound piRNAs sequenced in our previous study 36 . qRT-PCR quantifications of the selected piRNAs in low, medium, and high samples revealed that all 16 examined piRNAs were invariably upregulated along with the increase of cell density (Fig. 4E). Considered collectively, these results suggest that increasing BmN4 cell density globally activates the piRNA biogenesis mechanism via both primary and secondary pathways, resulting in the widespread accumulation of piRNAs in higher-density cells.
Cell density-dependent transposon silencing. As cell density regulates the abundance of piRNAs, which function to silence transposons, we investigated whether cell density affects the piRNA-mediated transposon silencing. To explore and identify the transposon species whose expressions are well regulated by piR-NAs in BmN4 cells, we knocked-down Siwi expression and analyzed the eight transposons (Yamato, Kimono, Ichiro, Yokohama, Noguchi, Yaocho, Kimutaku, and Nukegara) in which antisense strands produce abundant Siwi-bound piRNAs 23 . As in our previous study, RNAi-mediated silencing of Siwi reduced the expression levels of Siwi mRNA (Fig. 5A), Siwi protein (Fig. 5B), and Siwi-bound piR-1 (Fig. 5C). Among the eight transposons, Yamato and Kimono showed clear derepression upon the reduction of Siwi/piRNA complexes (Fig. 5D), indicating that Yamato and Kimono are the transposons whose expressions are well silenced by the piRNA pathway in BmN4 cells. The expression levels of both Yamato and Kimono decreased as cell density increased (Fig. 5E), which corresponds to the increase of piRNA abundance in increased cell density. These results suggest that cell density-dependent piRNA biogenesis and its resultant piRNA accumulation indeed have functional significance for the silencing of transposon expressions.

Discussion
Here we demonstrated that the piRNA pathway is intimately associated with cell density in Bombyx BmN4 cells.
As the cells are cultured with increasing cell density, Piwi proteins, piRNA biogenesis factors, and mature piRNAs are increasingly accumulated, whereas levels of a long piRNA precursor and transposons are reduced (Fig. 6). The consistent mRNA levels of Piwi proteins and piRNA biogenesis factors despite varying cell densities (Fig. 1D) suggest that their protein levels are regulated through a cell-density-linked post-transcriptional event. One of the possible post-transcriptional events would be the enhancement of protein stability by forming complexes in the perinuclear nuage-like granules, as the formation of the granule is also dependent on cell density (Fig. 2). One of the examined piRNA biogenesis factor, BmPapi, is localized on the outer membrane of the mitochondria 36 , yet increasingly accumulates in high-density cells. Therefore, cell density-dependent accumulation is a common    feature of all four of the examined piRNA biogenesis-related factors localizing in different locations in BmN4 cells (Fig. 1C). The increase of piRNA biogenesis factors could promote piRNA production from precursors, which could be the reason for the decreased levels of a piRNA precursor (Fig. 3C). As a result, mature piRNA levels are increased (Figs 3 and 4), eventually resulting in increased repression of transposon expression in the higher  density cells (Fig. 5E). Although the levels of miRNAs, as well as piRNAs, are increased in high-density cells, cell density-dependent regulatory pathway for piRNAs and miRNAs might be distinct, because, in human cancer cells, cell density regulates miRNA levels through Drosha activity but does not affect the levels of Argonatue proteins 44 .
The cell density-dependent regulation of the piRNA pathway was not caused by the changes in nutrients or other diffusible factors in the culturing medium (Fig. 1E). Although cell density is known to affect cell proliferation 41,47 , and indeed, the increased density of BmN4 cells enhanced cell proliferation (Fig. S5A), our cell cycle arrest experiments suggested that cell proliferation rate is not directly associated with the piRNA pathway ( Fig. S5B and C). Therefore, we reason that physical cell-cell contact could be the factor contributing to the regulation of the piRNA pathway. Cell-cell contact-dependent signals fulfill key and fundamental roles in animal development and cell differentiation 48,49 . For example, it is well known that E-cadherin-mediated cell-cell contact and subsequent maturation of adherens junctions with β-catenin mediate the Wnt signaling pathway 50,51 . Our results are in particular reminiscent of the regulation of Piwi expression through cell-cell communications in the Drosophila stem cell niche where the supporting cells exist to strictly control the maintenance and differentiation of germline stem cells 49 . In the niche, the Bruton's tyrosine kinase 29 A phosphorylates β-catenin under Wnt signaling pathway, which enhances Piwi protein expression to terminate germ cell proliferation 52 . Given these reports, our observations might not just represent an artificial phenomenon specific to BmN4 cells, but rather reflect the important regulatory mechanisms of the piRNA pathway in the context of germline development in organisms.
BmN4 cells are the only reported germ cells equipping the endogenous piRNA pathway, and they are therefore widely used as a cell line system to analyze piRNA biogenesis and function. Our findings indicate that cell density is a critical variable that should be closely monitored for accurate analyses of piRNA expression, biogenesis, and function in cell culture settings. Moreover, our findings suggest the utilization of cell culturing with different cell densities as a unique tool to monitor piRNA biogenesis. To date, the mechanism underlying the production of piRNA precursors and their processing into mature piRNAs have remained elusive. Although the use of various knockout or knockdown organisms with depleted expression of targeted piRNA biogenesis factors has greatly contributed to clarifying piRNA biogenesis and function, these studies always observed the "end point" resulting from the factor depletion, and they suffered limitations in monitoring the intermediate state of the phenomenon. As gradual regulation of the piRNA pathway is achieved by changing culturing cell density, RNA-seq and proteomic analyses of BmN4 cells with different cell densities might serve as a useful system for the analyses of piRNA precursors, piRNA biogenesis factors, and formation of perinuclear germ granules. Western blot. Western blot using BmN4 cell lysates was performed as described previously 36 . Cell lysates were prepared from identical number of the cells from respective density conditions in a lysis buffer containing 20 mM Tris-HCl pH 7.4, 200 mM NaCl, 2.5 mM MgCl 2 , 0.5% NP-40, 0.1% Triton X-100, and complete protease inhibitor (Roche Diagnostics). After quantifying protein concentration of the lysates, 20 μg of each lysate was loaded onto SDS-PAGE. The following antibodies were used: S213 anti-Siwi 36 , anti-BmAgo3 23 , anti-Tom20 (Santa Cruz Biotechnology), anti-Hsp60 (Cell Signaling), anti-β-actin (Abcam), anti-BmPapi 10 , and BmVasa571 anti-BmVasa 53 . The anti-BmVasa is able to specifically recognize BmVasa in Western blot and immunofluorescence ( Supplementary Fig. S1). The Western blot band intensities showed clear linearity to the amount of lysate input (2.5-20 μg), suggesting the quantification ability of the Western blot analyses (Supplementary Fig. S2).

Methods
Immunofluorescence and confocal microscope. BmN4 cells were plated on a slide glass chamber (Lab-Tek) with differing densities and incubated for 30 h. Immunofluorescence staining was performed as described previously 36 using anti-Siwi (diluted 1:2000), anti-BmAgo3 (1:5000), and anti-BmVasa (1:1000) as primary antibodies. Alexa Fluor 488 goat anti-rabbit IgG (Life Technologies) was used as a secondary antibody. After DNA counterstaining with ProLong Gold Antifade Reagent with DAPI (Life Technologies), images shown in Fig. 2A and Supplementary Fig. S1B were acquired using a Leica SP5 confocal microscope and a Nikon Eclipse Ti-U microscope, respectively. In Fig. 2B, signal intensities of each protein were quantified using ImageJ software. The color images were divided to two images (target protein and DAPI) by running split channel, and, after running despeckle to remove noises, segmentations were performed by setting the following threshold values: Siwi, 100-255; BmAgo3, 50-255; BmVasa, 100-255; and DAPI: 30-255. The integrated intensities of target proteins were then measured and normalized to those of DAPI.
Next-generation sequencing of RNAs. The total RNA was extracted from BmN4 cells after 30 h culture with the starting cell densities of 0.6 × 10 3 (low), 3.0 × 10 3 (medium), or 6.0 × 10 3 (high) cells/cm 2 . RNAs with lengths under 150 nt were gel-purified and subjected to cDNA amplifications and next-generation sequencings using 5500xl SOLiD System (Life Technologies) at the Cancer Genomics Laboratory of the Sidney Kimmel Cancer Center of Thomas Jefferson University.
Bioinformatics analyses. RNA seq data from BmN4 cells with low-, medium-, and high-densities contained 117,552,644, 144,877,411, and 152,257,139 raw reads, respectively, and can be found publicly at NCBI's Sequence Read Archive (accession no. SRP104077). Before mapping, we used the cutadapt tool (http://dx.doi. org/10.14806/ej.17.1.200; http://journal.embnet.org/index.php/embnetjournal/article/view/200) to perform quality check and adapter trimming. The processed sequences were non-uniquely mapped to the B. mori genome extracted from SilkDB v2.0 (http://www.silkdb.org/silkdb/) by SHRiMP2 55 , allowing a 4% mismatch rate. No insertions or deletions were permitted and reads that mapped to >10,000 places were removed from subsequent analyses. For nucleotide composition analysis of piRNAs, 24-30-nt reads were non-uniquely mapped to 1,811 B. mori transposons 56 , and the mapped reads were applied to FastQC. Ping-pong analysis was performed by calculating distances between 5′-ends of piRNAs across opposite genomic strand as described previously 36 . The reads uniquely mapped to 1,811 B. mori transposons were used for the ping-pong analysis.