Abstract
CD95/Fas ligand binds to the death receptor CD95/Fas to induce apoptosis in sensitive cells. We previously reported the CD95L mRNA is enriched in sequences that, when converted to si/shRNAs, are toxic to cells (Putzbach et al., 2017). These si/shRNAs kill all cancer cells through a RNAi off-target effect by targeting critical survival genes. We now report expression of full-length CD95L mRNA, itself, is highly toxic to cells and induces a similar form of cell death. We demonstrate that small RNAs derived from CD95L are loaded into the RNA induced silencing complex (RISC) and that the RISC is required for the toxicity. Drosha and Dicer knock-out cells are highly sensitive to this toxicity, suggesting that processing of CD95L mRNA into small toxic RNAs is independent of both Dicer and Drosha. The data provide evidence that a higher vertebrate transgene can be processed to RNAi-active small RNAs that elicit cellular responses.
Introduction
Activation of CD95/Fas through interaction with its cognate ligand CD95L or receptor-activating antibodies induces apoptosis in sensitive cells (Suda, Takahashi, Golstein, & Nagata, 1993). Virtually all research on CD95 and CD95L has focused on the physical interaction between the two proteins and the subsequent protein-based signaling cascades (Algeciras-Schimnich et al., 2002; Fu et al., 2016; Nisihara et al., 2001; Schneider et al., 1997). However, we have recently shown that the mRNA of CD95 and CD95L harbor sequences that when converted into small interfering (si) or short hairpin (sh)RNAs, cause massive and robust toxicity in all tested cancer cells. These CD95/CD95L-derived si/shRNAs target a network of survival genes, resulting in the simultaneous activation of multiple cell death pathways through RNA interference (RNAi) in a process we called DISE (Death Induced by Survival gene Elimination) (Putzbach et al., 2017). We determined that for an si/shRNA to elicit this form of toxicity, only positions 2-7 of the guide strand, the 6mer seed sequence, are required (Putzbach et al., 2017). More recently, a screen of all 4096 6mer seeds revealed that optimal 6mer seed toxicity requires G-rich seeds targeting C-rich regions in the 3′UTRs of survival genes (Gao et al., 2018).
In this report, we show that expression of the CD95L mRNA, itself, is toxic to cells even without prior conversion to siRNAs. This toxicity is independent of the full-length CD95L protein or expression of the CD95 receptor and resembles DISE. The toxicity involves RNAi, and multiple small RNAs generated within cells from the mRNA of CD95L are loaded into the RNA-induced Silencing Complex (RISC), the key mediator of RNAi (Liu et al., 2004).
Results
CD95L mRNA is toxic to cells
By testing every possible shRNA derived from the CD95L open reading frame (ORF) or its 3′UTR, we recently found a high enrichment of toxic si/shRNAs derived from the CD95L ORF (Putzbach et al., 2017). Most recently we determined that the 6mer seed toxicity observed in many si/shRNAs is due to their nucleotide composition, with G-rich seeds being the most toxic (Gao et al., 2018). When reanalyzing the CD95L ORF-derived shRNAs, we found a significant correlation between the toxicity of the most toxic CD95L-derived shRNAs (Putzbach et al., 2017) and the seed toxicity of the same 6mer seed we recently determined in a screen of all 4096 6mer seeds (Gao et al., 2018). This suggests that CD95L-derived shRNAs kill cancer cells mainly through 6mer seed toxicity.
We therefore wondered whether expression of the CD95L ORF mRNA—without pre-processing into artificial siRNAs—would be toxic to cells. Expression of CD95L protein in most cells kills through induction of apoptosis. Consequently, expressing CD95L in HeyA8 cells, which are highly sensitive to CD95 mediated apoptosis, killed cells within a few hours after infection with a lentivirus encoding CD95L (Figure 1A, left panel). Interestingly, severe growth reduction was seen without any signs of apoptosis (not shown) when a CD95L mutant, unable to bind CD95, was expressed (CD95LMUT in Figure 1A, left panel). This mutant carries a Y218R point mutation, which prevents the CD95L protein from binding to CD95 (Schneider et al., 1997), and is expressed at a similar level to wild type (wt) CD95L (Figure 1B). To prevent the CD95L mRNA from producing full-length CD95L protein, we also introduced a premature stop codon right after the start codon in the CD95LMUT vector (CD95LMUTNP). This construct (containing 4 point mutations and confirmed to produce mRNA with no detectable full-length CD95L protein, Figure 1B) was equally active in reducing the growth of HeyA8 cells when compared to the CD95LMUT vector (Figure 1A, left panel). This result suggested that the CD95L mRNA could be toxic to HeyA8 cells without the CD95L protein inducing apoptosis. This was confirmed by expressing the three CD95L constructs in the presence of the oligo-caspase inhibitor zVAD-fmk (Figure 1A, center panel). With suppressed apoptosis, all three constructs were now equally toxic to HeyA8 cells. Finally, we tested a HeyA8 CD95 k.o. clone confirmed to express no CD95 protein (Putzbach et al., 2017). In these cells, without the addition of zVAD-fmk, wt CD95L and CD95LMUTNP were again equally active in severely reducing the growth of the cells (Figure 1A, right panel). Together, these data suggested it is the CD95L mRNA that killed the cells. Cell death was confirmed by quantifying nuclear fragmentation (Figure 1C). We also detected a significant increase of ROS in cells expressing CD95LMUTNP (Figure 1D), which is a characteristic feature of DISE (Hadji et al., 2014; Patel & Peter, 2017). To exclude the possibility that truncated CD95 protein or any part of the CD95 mRNA would play a role in this toxicity, we deleted the CD95 gene in MCF-7 cells (Figure 1 -figure supplement 1A-E). Overexpression of wild-type CD95L killed clone FA4 cells, which harbor a complete homozygous deletion of the entire CD95 gene, as well as CD95 protein k.o. clone #21 cells that retain some truncated mRNA expression (Figure 1E). To further distinguish the activities of the CD95L mRNA from CD95L protein, we generated a CD95L expression construct in which we introduced 308 (out of 846 nucleotides) silent mutations (CD95L SIL) (Figure 1 - figure supplement 2A). The activity of this mutant construct to negatively affect cell growth of CD95 k.o. HeyA8 cells was compared to two independently cloned wt CD95L constructs (WT1 and WT2 in Figure 1 - figure supplement 2B). The mutant SIL construct was equally effective in suppressing cell growth and produced about the same amount of CD95L mRNA (Figure 1 - figure supplement 2C). However, the SIL construct only produced about 12 % of WT CD95L protein (Figure 1 - figure supplement 2C), again supporting the observation that it is the CD95L RNA and not the protein that elicits toxicity.
CD95L mRNA kills cells through DISE
After infection with CD95L, CD95 k.o. HeyA8 cells exhibited morphological changes strikingly similar to the changes seen in wt HeyA8 cells after introduction of a CD95L-derived shRNA (shL3) (Figure 2A, Video 1-4) suggesting the cells died through a similar mechanism. To determine the cause of cell death induced by CD95L mRNA in HeyA8 CD95 k.o. cells molecularly, we performed an RNA-Seq analysis. We found that expression of CD95L caused preferential downregulation of critical survival genes and not of nonsurvival genes in a control set (Figure 2B). In addition, cell death induced by CD95L mRNA resulted in a substantial loss of 11 of the 12 histones detected to be downregulated in cells treated with CD95 and CD95L-derived sh/siRNAs (Figure 2C). Loss of histons is an early event during DISE (Putzbach et al., 2017). A Metascape analysis demonstrated that nucleosome assembly, regulation of mitosis, and genes consistent with the involvement of histones were among the most significantly downregulated RNAs across all cells in which DISE was induced by any of the four sh/siRNAs or by the expression of CD95L mRNA (Figure 2D). This suggests that CD95L mRNA kills cells in the same way as CD95/L-derived si/shRNAs.
CD95L mRNA kills cells through RNAi
Given our previous work on CD95L-derived si/shRNA toxicity, we hypothesized that CD95L mRNA kills cells through an RNAi-based mechanism—perhaps by being processed into small RNAs that incorporate into the RISC. Drosha k.o. cells lacking the majority of endogenous miRNAs, but retaining expression of Ago proteins, were shown to be hypersensitive to DISE induced by si-and shRNAs (Putzbach et al., 2017). We interpreted this effect as being caused by an increased pool of unoccupied RNAi machinery caused by the absence of most miRNAs. Drosha k.o. cells were also hypersensitive to the expression of CD95LMUTNP (Figure 3A, p=0.014, according to a polynomial fitting model); Virtually all cells died (insert in Figure 3A). To directly determine whether the RISC is involved in the toxicity, we introduced CD95L into CD95 k.o. HeyA8 cells after knocking down AGO2 (Figure 3B). Knock down of AGO2 was efficient (insert in Figure 3B). Toxicity elicited by CD95L was severely blunted following AGO2 knockdown, suggesting that AGO2 was required for CD95L mRNA to be toxic.
To test the hypothesis that Drosha k.o. cells were more sensitive because their RISC was not occupied by large amounts of nontoxic miRNAs and to determine whether CD95L mRNA could give rise to small RNAs that incorporate into the RISC, we pulled down AGO1-4-associated RNAs and analyzed their composition in wt and Drosha k.o. cells after expressing the CD95LMUTNP mRNA. For the pull-down, we used a peptide—derived from the AGO-binding partner GW182—recently described to bind to all four Ago proteins (Hauptmann et al., 2015). As expected in wt HCT116 cells, large amounts of small RNAs (19-23nt in length) were detected bound to the Ago proteins (Figure 3C). Both AGO1 and AGO2 were efficiently pulled down. In contrast, in the Drosha k.o. cells, which cannot generate canonical miRNAs, only a low amount of small RNAs was detected, confirming the absence of miRNAs in the RISC. Surprisingly, the amount of pulled down Ago proteins was severely reduced despite the fact these Drosha k.o. cells express comparable levels of AGO2 (Putzbach et al., 2017). This suggests the peptide did not have access to the Ago proteins in Drosha k.o. cells, presumably because it only binds to Ago proteins complexed with RNA as recently shown (Elkayam et al., 2017).
The analysis of all Ago-bound RNAs showed that in the wt cells, >98.4% of bound RNAs were miRNAs. In contrast, only 34% of bound RNAs were miRNAs in Drosha k.o. cells (Figure 3D and data not shown). These include miRNAs that are processed independently of Drosha such as miR-320a (Kim, Kim, & Kim, 2016). Consistently, this miRNA became a major RNA species bound to Ago proteins in Drosha k.o. cells (Figure 3D). In both wt and Drosha k.o. cells, a significant increase in CD95L-derived small RNAs bound to the Ago proteins was detected compared to cells infected with pLenti empty vector. They corresponded to 0.0006% and 0.043% of all the Ago-bound RNAs in the wt cells and Drosha k.o. cells, respectively. Toxicity of CD95L mRNA was, therefore, not due to overloading the RISC. In the absence of most miRNAs, the total amount of RNAs bound to Ago proteins in the Drosha k.o. cells was roughly 10% of the amount bound to Ago in wt cells (Figure 3D). The reduction of Ago-bound miRNAs in Drosha k.o. cells (Figure 3E, top row) was paralleled by a substantial increase in binding of other small RNAs to the Ago proteins (Figure 3E, bottom row). Interestingly, the amount of Ago-bound CD95L-derived small RNAs was >100 times higher in the Drosha k.o. cells compared to the wt cells (red columns in Figure 3E). These data support our hypothesis that Drosha k.o. cells are more sensitive to CD95L mRNA-mediated toxicity due to their ability to take up more toxic small CD95L-derived RNAs into the RISC in the absence of most miRNAs.
CD95L ORF is degraded into small RNA fragments that are then loaded into the RISC
Interestingly, not only did Ago proteins in Drosha k.o. cells bind much more CD95L-derived small RNAs than in the wt cells, but also the peak length of the most abundant Ago-bound RNA species increased from 20 to 23 nt (Figure 4A, top panel). To determine the sites within the CD95L mRNA that gave rise to small Ago-bound RNAs, we aligned all small Ago-bound RNAs detected in all conditions to the CD95L ORF sequence (Figure 4B and C). We identified 22 regions in the CD95L ORF that gave rise to small RNAs that could be bound by Ago proteins (Figure 4B). To determine whether these small RNAs were formed in the cytosol and then loaded into the RISC, we also aligned all small RNAs in the total RNA fraction isolated from CD95LMUTNP expressing HCT116 Drosha k.o. cells with CD95L (Figure 4C). Interestingly, very similar regions of small RNAs were found. Moreover, the mean as well as the peak of the distribution of the read lengths of small RNAs bound to Ago proteins was smaller than in the total small RNAs fraction (Figure 4A, center panel), suggesting these fragments were trimmed to the appropriate length either right before they are loaded into the RISC or by the RISC itself. This was most obvious for the small RNAs in cluster 3 (Figure 4B and C). We also noticed that certain small RNAs were more abundant in the Ago-bound fraction when compared to total RNA relative to all other RNAs. To determine whether this type of processing was specific for HCT116 Drosha k.o. cells, we analyzed the Ago-bound small CD95L-derived RNAs in HeyA8 CD95 k.o. cells after expression of wt CD95L (Figure 4D) and compared them with the total RNA fraction (Figure 4E). While we found fewer CD95L-derived reads in these cells, the general location of some of the read clusters overlapped with the ones found in the Drosha k.o. cells and again both the mean and peak of the distribution of RNA lengths was smaller in the Ago-bound fraction versus the total RNA fraction (Figure 4A, bottom panel). Together, these data suggest that CD95L mRNA can be processed into smaller RNA fragments, which are then trimmed to a length appropriate for incorporation into the RISC.
Our data suggest that the CD95L mRNA, when overexpressed, is toxic to cells due to the formation of Ago-bound small RNAs that are incorporated into the RISC and kill cells through RNAi. This process is independent of Drosha. To determine whether Dicer is required for either processing of CD95L mRNA or loading the small RNAs into the RISC, we expressed CD95LMUTNP in wt and Dicer k.o. HCT116 cells (Figure 4F). Dicer k.o. cells were still sensitive to toxicity induced by CD95L mRNA expression, suggesting the toxicity of the CD95L mRNA does not require the processing by either Drosha or Dicer. Using custom real-time qPCR primers designed to specifically detect the small RNAs from clusters 8 and 21, we detected, in both wt and Dicer k.o. cells over-expressing CD95LMUTNP, fragments from these clusters (Figure 4G), demonstrating that Dicer is not involved in processing CD95L mRNA.
All the reported small RNAs derived from CD95L corresponded to the sense strand of the expressed mRNA, raising the question of how they could be processed into double-stranded siRNAs in the absence of an antisense strand. To get a preliminary answer to this question, we subjected the CD95L ORF mRNA sequence to a secondary structure prediction (Figure 4 - figure supplement 1A). According to this analysis, the CD95L ORF mRNA forms a tightly folded structure with many of the small RNAs of the 22 clusters juxtaposing each other in stem-like structures creating regions of significant complementarity. These may provide the duplexes needed to be processed and loaded into the RISC. Interestingly, many of the juxtaposing reads were found in duplex structures with 3′ overhangs. Three of these oligonucleotides (derived from clusters 7, 15 and 22) when expressed as siRNAs were toxic to HeyA8, H460, M565 and 3LL cells (Figure 4 - figure supplement 1B).
In summary, our data suggest that si- and/or shRNAs with certain seed sequences as they are present in CD95 and CD95L and the entire CD95L ORF are toxic to cancer cells. The CD95L mRNA is broken down into small RNA-active fragments that are loaded into the RISC and then target critical survival genes. This results in cell death through 6mer seed toxicity. The process is independent of both Drosha and Dicer. Finally, the data suggest that a high miRNA content, by “filling up” the RISC, might render cells less sensitive to this form of cell death.
Discussion
We recently reported a novel form of cell death that was observed after expression of si/shRNAs designed from the sequences of CD95/CD95L mRNA (Putzbach et al., 2017). More recently we described that cells die from a loss of multiple survival genes through a mechanism we call 6mer seed toxicity (Gao et al., 2018). The most toxic si/shRNAs derived from CD95 or CD95L were found in the ORF of CD95L (Putzbach et al., 2017). This pointed toward the CD95L mRNA, itself being toxic.
We now show that expression of full-length CD95L mRNA triggers toxicity that is independent of the protein product and canonical apoptosis. This is intriguing considering previous studies showed transgenic expression of CD95L using viruses killed multiple cancer cells that were completely resistant to CD95 mediated apoptosis after addition of agonist anti- CD95 antibodies (ElOjeimy et al., 2006; Hyer, Voelkel-Johnson, Rubinchik, Dong, & Norris, 2000; Sudarshan et al., 2005; Sun et al., 2012). These results were interpreted as intracellular CD95L triggering apoptosis. However, we now provide an alternate explanation—namely, both the CD95L protein and mRNA are toxic to cells through distinct mechanisms. The protein induces apoptosis, and the mRNA induces toxicity through an RNAi-based mechanism.
We demonstrate that Dicer and Drosha are not involved in generating the Ago-bound CD95L-derived fragments but there are several candidate RNases that are capable of processing mRNAs. Given the differences in length distribution between the cytosolic versus Ago-bound RNA fragments, it is likely the released CD95L-derived fragment intermediates are incorporated into the RISC and then trimmed to the appropriate length by Ago. Indeed, a similar mechanism is known to occur during the maturation of the erythropoietic miR-451, where the pre-miRNA is first cleaved by AGO2 and then trimmed at the 3’ end to the final mature form by the exoribonuclease PARN (Yoda et al., 2013). Furthermore, a similar process occurs with the recently identified class of Ago-bound RNAs called agotrons (Hansen et al., 2016), which consist of an excised intron loaded into the RISC in a manner independent of Drosha or Dicer pre-processing. After trimmed to the appropriate size, the guide RNAs in complex with the RISC can regulate gene expression through RNAi.
Our data provide the first evidence of an overexpressed cDNA to be toxic via an RNAi-dependent mechanism. It was first shown in plants that overexpressed transgenes can be converted into RNAi active short RNA sequences (Hamilton & Baulcombe, 1999). Our data on the effects of overexpressed CD95L RNA, while mechanistically distinct from what was reported in plants, maybe the first example of transgene determining cell fate through the RNAi mechanism in mammalian cells.
A major question that arises from our data is whether CD95L mRNA is toxic in vivo. We and others have noticed upregulation of CD95L in multiple stress-related conditions such as after treatment with chemotherapy ((Friesen, Fulda, & Debatin, 1999) and data not shown). While the amount of CD95L mRNA and the level of upregulation alone may not be enough to be toxic, it could be the combination of multiple RNA fragments, derived from multiple different mRNAs that are generated to kill cells (Putzbach et al., 2018). We view CD95L as just one of many RNAs that have this activity. Indeed, it is unlikely CD95L is the only gene whose mRNA is toxic to cells, as this mRNA-based level of toxicity would be redundant with the potent killing capacity of the CD95L protein. Also, upregulating an mRNA that, by itself, could decimate the cells that would otherwise need to upregulate that mRNA to carry out their biological function in the first place, such as activated T cells upregulating CD95L to mount an immune response, would be self-defeating. Therefore, nature likely distributed this mRNA-based toxicity-inducing capacity over many genes in the genome to prevent activating it when any one of those genes is upregulated during specific cellular processes. It is more likely there exists an entire network of these genes that can release toxic small RNAs when the appropriate stimulus is encountered. Consistent with this hypothesis we recently identified other genes that contain sequences that when converted to shRNAs kill cancer cells through 6mer seed toxicity (Patel & Peter, 2017). Future work will be aimed at identifying additional genes and the mechanism through which they are processed and under what conditions to kill cells.
Materials and methods
Reagents and antibodies
All reagents and antibodies were described previously (Putzbach et al., 2017) except those referenced in the following paragraphs.
Cell lines
HeyA8 (RRID:CVCL_8878) and HeyA8 CD95 knock-out cells, HCT116 (ATCC #CCL-247; RRID:CVCL_0291) and HCT116 Drosha knock-out and Dicer knock-out cells, MCF-7 cells (ATCC #HTB-22; RRID:CVCL_0031), and 293T (ATCC #CRL-3216; RRID:CVCL_0063) cells were cultured as described previously (Putzbach et al., 2017). The MCF-7 CD95 knock-out and deletion cells were cultured in RPMI 1640 medium (Cellgro #10-040-CM), 10% heat-inactivated FBS (Sigma-Aldrich), 1% L-glutamine (Mediatech Inc), and 1% penicillin/streptomycin (Mediatech Inc). H460 (ATCC #HTB-177; RRID:CVCL_0459) cells were cultured in RPMI1640 medium (Cellgro Cat#10-040) supplemented with 10% FBS (Sigma Cat#14009C) and 1% L-Glutamine (Corning Cat#25-005). 3LL cells (ATCC #CRL-1642; RRID:CVCL_4358) were cultured in DMEM medium (Gibco Cat#12430054) supplemented with 10% FBS and 1% L-Glutamine. Mouse hepatocellular carcinoma cells M565 cells were described previously (Ceppi et al., 2014) and cultured in DMEM/F12 (Gibco Cat#11330) supplemented with 10% FBS, 1% L-Glutamine and ITS (Corning #25-800-CR). All cell lines were authenticated using STR profiling and tested monthly for mycoplasm using PlasmoTest (Invitrogen).
Plasmids and constructs
The pLenti-CD95L was synthesized by sub-cloning an insert containing the CD95L ORF (NM_000639.2; synthesized by IDT as minigene with flanking 5’ NheI RE site and 3’ XhoI RE sites in pIDTblue vector) into the pLenti-GIII-CMV-RFP-2A-Puro vector (ABM Inc). The insert and the backbone were digested with NheI (NEB #R0131) and XhoI (NEB #R0146) restriction enzymes. Subsequent ligation with T4 DNA ligase created the pLenti-CD95L vector. The pLenti-CD95LMUT vector was created by sub-cloning a CD95L cDNA insert with 2 nucleotide substitutions in codon 218 (TAT -> CGT) resulting in replacement of tyrosine for arginine, which has been described to inhibit binding to CD95 (Schneider et al., 1997) into the pLenti-GIII-CMV-RFP-2A-Puro vector. The pLenti-CD95LMUTNP vector was created by inserting a CD95L ORF cDNA sequence containing both the Y218R mutation and a single nucleotide substitution at the second codon (CAG -> TAG), resulting in a premature stop codon right after the start codon, into the pLenti-GIII-CMV-RFP-2A-Puro vector. The pLenti-CD95LSIL was created by sub-cloning a mutant CD95L ORF cDNA sequence with codons synonymously mutated (Figure 1 - figure supplement 2A) to the next most highly utilized codon in human cells (exceptions were made within the proline rich domain to meet gene synthesis design criteria.) into the pLenti-GIII-CMV-RFP-2A-Puro vector.
Overexpression of CD95L cDNAs
All lentiviral constructs were generated in 293T cells as described previously (Putzbach et al., 2017). HeyA8 and MCF-7 (and all derivative cell lines) cells overexpressing wild type CD95L and mutant CD95L cDNAs were generated by seeding cells at 100,000 cells per well in a 6-well plate and infecting cells with lentivirus generated in 293T cells (500 µl viral supernatant per well) with 8 μg/ml Polybrene. Media was changed next day. Selection was started either during the evening of the same day or on following day with 3 μg/ml puromycin. HCT116, HCT116 Drosha knockout, and HCT116 Dicer knockout cells (Kim et al., 2016) overexpressing CD95L cDNAs were generated by seeding cells at 100,000 cells per well in a 24-well plate or 500,000 cells per well in a 6-well plate and infecting cells with lentivirus generated in 293T cells (100 μl virus per 24-well or 500 μl per 6-well) in the presence of 8 μg/ml Polybrene. Media was changed the next day, and cells were selected with 3 μg/ml puromycin the following day. Infection with empty pLenti was always included as a control.
To assess toxicity of overexpressing CD95L cDNAs, cells infected with these constructs were plated in on a 96-well plate 1 day after selection in the presence of puromycin (uninfected cells were all dead after 1 day in presence of puromycin); Cell confluency was assessed over time using the IncuCyte as described previously (Putzbach et al., 2017).
To assess overexpression of CD95L cDNAs in apoptosis-sensitive HeyA8 cells in Figure 1A, infection with CD95L lentiviruses were done in 96-well plate using 50 µl of virus in the presence of 20 µM zVAD-fmk (Sigma-Aldrich #V116) and 8 μg/ml Polybrene; media was changed next day in the presence of 20 µM zVAD-fmk; 3 μg/ml puromycin was added the following day. Infection with the CD95L constructs for the RT-qPCR and Western blot in Figure 1B were done in a 6-well plate in the presence of 20 µM zVAD-fmk.
For the experiment in Figure 3B, HeyA8 CD95 knock-out cells were reverse transfected in a 6-well plate; 100,000 cells were plated in wells with either the On-TargetPlus non-targeting siRNA (Dharmacon #D-001810-10) or siAGO2 pool (Dharmacon ##L-004639-00-0005) at 25 nM complexed with 1 μl RNAiMax. After ~24 hrs, the cells were infected with either pLenti or pLenti-CD95L (500 μl of viral supernatant [25% of total volume]). Next day, media was replaced, and cells were expanded to 10 cm plates. The following day, 3 μg/ml puromycin was added. When puromycin selection was complete (one day later), the 750-1,500 cells were plated per well in a 96-well plate and put in the IncuCyte machine to assess cell confluency over time.
CRISPR deletions
We co-transfected a Cas9-expressing plasmid (Jinek et al., 2013) and two gRNAs that target upstream and downstream to delete an entire section of DNA as described previously (Putzbach et al., 2017). The gRNA scaffold was used as described (Mali et al., 2013). The gRNAs were designed using the algorithm found at http://crispr.mit.edu; only gRNAs with a score above 50 were considered.
A deletion of 227 nucleotides in exon 4 of CD95 in MCF-7 cells (ΔshR6, clone #21) was generated using gRNAs described previously (Putzbach et al., 2017). Deletion of this site results in a frame-shift mutation that causes a protein-level knock-out (Putzbach et al., 2017). PCR with flanking external primers (Fr: 5’-GGTGTCATGCTGTGACTGTTG-3’ and Rev: 5’- TTTAGCTTAAGTGGCCAGCAA-3’) and internal primers (Fr primer and the internal Rev primer 5’-AAGTTGGTTTACATCTGCAC-3’) was used to screen for single cell clones that harbor a homozygous deletion.
The two sequences targeted by the flanking gRNAs for the deletion of the entire CD95 gene were 5’-GTCAGGGTTCGTTGCACAAA-3’ and 5’-TGCTTCTTGGATCCCTTAGA-3’. For detection of the CD95 gene deletion, the flanking external primers were 5’-TGTTTAATATAGCTGGGGCTATGC-3’ (Fr primer) and 5’-TGGGACTCATGGGTTAAATAGAAT-3’ (Rev primer), and the internal reverse primer was 5’-GACCAGTCTTCTCATTTCAGAGGT-3’. After screening the clones, Sanger sequencing was performed to confirm the proper deletion had occurred.
Real-Time quantitative PCR
The relative expression of specific mRNAs was quantified as described previously (Gao et al., 2018). The primer/probes purchased from ThermoFisher Scientific were GAPDH (Hs00266705_g1), human CD95L (Hs00181226_g1 and Hs00181225_m1), human CD95 (Hs00531110_m1 and Hs00236330_m1), and a custom primer/probe to detect the CD95LSIL mRNA (designed using the Thermofisher Scientific custom design tool; assay ID: APNKTUD).
Custom RT-qPCR probes designed to specifically detect small RNA species were used to detect CD95L fragments in Figure 4G. These probes were designed using ThermoFisher’s Custom TaqMan Small RNA Assay Design Tool (https://www.thermofisher.com/order/custom-genomic-products/tools/small-rna/) to target the cluster 8 sequence (5’-AAGGAGCTGGCAGAACTCCGAGA-3’) and the cluster 21 sequence (5’-TCAACGTATCTGAGCTCTCTC-3’). Detection of these fragments involves a two-step amplification protocol used to detect microRNAs. In the first step, the High-Capacity cDNA reverse transcription kit is used to selectively reverse transcribe the two clusters to be quantified using specific primers and 20 nM RNA input following the manufacturer’s protocol. The cDNA is diluted 1:5. The qPCR reaction mixture is composed of the diluted cDNA, the custom probes, and the Taqman Universal PCR Master Mix (Applied Biosystems #43240018). Reactions were performed in triplicate. Ct values were determined using the Applied Biosystems 7500 Real Time PCR system with a thermocycle profile of 50°C for two min (step one), 95°C for 10 min (step two), and then 40 cycles of 95°C for 15 s (step three) and 60°C for 1 min (step four). The ΔΔCt values between the small RNA of interest and the control were calculated to determine relative abundance of the small RNA. Samples were normalized to Z30 (ThermoFisher Scientific #4427975).
Western blot analysis
Detection of human CD95, CD95L and Ago proteins was done via Western blot as described previously (Putzbach et al., 2017).
CD95 surface staining
Flow cytometry was used to quantify the level of membrane-localized CD95 as described previously (Putzbach et al., 2017).
Cell death quantification (DNA fragmentation) and ROS production
The percent of subG1 nuclei (fragmented DNA) was determined by PI staining/flow cytometry as described previously (Putzbach et al., 2017). ROS production was quantified using the cell-permeable indicator 2′,7′-dichlorodihydrofluorescein diacetate (ThermoFisher Scientific #D399) as previously described (Hadji et al., 2014).
Assessing cell growth and fluorescence over time
After treatment/infection, cells were seeded in a 96-well plate at least in triplicate. Images were captured at indicated time points using an IncuCyte ZOOM live cell imaging system (Essen BioScience) with a 10x objective lens. Percent confluence and total fluorescent integrated intensity was calculated using the IncuCyte ZOOM software (version 2015A).
Infection of cells for Ago-pull down and small RNA-Seq analysis
HeyA8 ΔshR6 clone #11 cells were seeded at 75,000 cells per well on 6-well plates, and the HCT116 and HCT116 Drosha knock-out cells were both seeded at 500,000 per well on 6-well plates. The HeyA8 ΔshR6 clone #11 cells were infected with 0.5 mL of empty pLenti or pLenti-CD95L-WT viral supernatant per well. The HCT116 and HCT116 Drosha knockout cells were infected with 0.5 mL empty pLenti or pLenti-CD95LMUTNP viral supernatant per well. Media was changed the next day and the cells were pooled and expanded to multiple 15 cm dishes. Selection with 3 µg/mL puromycin began the following day. The next day, the HeyA8 ΔshR6 clone #11 infected cells were seeded at 600,000 cells per dish in multiple 15 cm dishes; the HCT116 and HCT116 Drosha knock-out cells were seeded at 5 million cells per dish in multiple 15 cm dishes. Two days later, each of the samples was pelleted and split in two: one pellet was lysed and processed for small RNA sequencing, and the other pellet was flash frozen in liquid nitrogen. The pellets were stored at −80°C until they could be used for the Ago pull-down experiment. The purpose of splitting the sample was so that we could compare the total cellular pool of small RNAs to the fraction that was bound to the RISC. This way, the processing CD95L-derived fragments from the full-length mRNA in the cytosol to the final mature RISC-bound form could be mapped. This was all done in duplicate.
RNA-Seq analysis
Total RNA was isolated using the miRNeasy Mini Kit (Qiagen, #74004) following the manufacturer’s instructions. An on-column digestion step using the RNase-free DNase Set (Qiagen #79254) was included. Both small and large mRNA libraries were generated and sequenced as described previously (Putzbach et al., 2017). Reads were trimmed with TrimGalore and then aligned to the hg38 assembly of the human genome with Tophat. Raw read counts were assigned to genes using HTSeq and differential gene expression was analyzed with the R Bioconductor EdgeR package (Robinson, McCarthy, & Smyth, 2010).
Ago pull down and RNA-Seq analysis of bound small RNAs
Cell pellets were harvest at 50 hours after plating (122 hours after infection) and were flash frozen in liquid nitrogen. The pellets were stored at −80°C until ready for further processing. Between 10 and 25 x 106 cells were lysed in NP40 lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 2 mM EDTA, 1% (v/v) NP40, supplemented with phosphatase inhibitors) on ice for 15 minutes. The lysate was sonicated 3 times for 30 s at 60% amplitude (Sonics, VCX130) and cleared by centrifugation at 12,000g for 20 minutes. AGO1-4 were pulled down by using 500 μg of Flag-GST-T6B peptide (Hauptmann et al., 2015) and with 60 μl anti-Flag M2 magnetic beads (Sigma-Aldrich) for 2 hrs at 4°C. The pull-down was washed 3 times in NP40 lysis buffer. During the last wash, 10% of beads were removed and incubated at 95°C for 5 minutes in 2x SDS-PAGE sample buffer. Samples were run on a 4-12% SDS-PAGE and transferred to nitrocellulose membrane. The pull-down efficiency was determined by immunoblotting against AGO1 (Cell Signaling #5053; RRID:AB_10695871 and Abcam #98056; RRID:AB_10680548) and AGO2 (Abcam #32381; RRID:AB_867543). To the remaining beads 500 µl TRIzol reagent were added and the RNA extracted according to the manufacturer’s instructions. The RNA pellet was diluted in 20 µl of water. The sample was split, and half of the sample was dephosphorylated with 0.5 U/µl of CIP alkaline phosphatase at 37°C for 15 min and subsequently radiolabeled with 0.5 µCi γ-32P-ATP and 1 U/µl of T4 PNK kinase for 20 min at 37°C. The AGO1-4 interacting RNAs were visualized on a 15% urea-PAGE. The remaining RNA was taken through a small RNA library preparation as previously described (Hafner et al., 2012). Briefly, RNA was ligated with 3’ adenylated adapters and separated on a 15% denaturing urea-PAGE. The RNA corresponding to insert size of 19-35 nt was eluted from the gel, ethanol precipitated followed by 5’ adapter ligation. The samples were separated on a 12% Urea-PAGE and extracted from the gel. Reverse transcription was performed using Superscript III reverse transcriptase and the cDNA amplified by PCR. The cDNA was sequenced on Illumina HiSeq 3000. Adapter sequences: Adapter 1 – NNTGACTGTGGAATTCTCGGGTGCCAAGG; Adapter 2 – NNACACTCTGGAATTCTCGGGTGCCAAGG, Adapter 3 – NNACAGAGTGGAATTCTCGGGTGCCAAGG, Adapter 4 – NNGCGATATGGAATTCTCGGGTGCCAAGG, Adapter 47 – NNTCTGTGTGGAATTCTCGGGTGCCAAGG, Adapter 48 – NNCAGCATTGGAATTCTCGGGTGCCAAGG, Adapter 49 – NNATAGTATGGAATTCTCGGGTGCCAAGG, Adapter 50 – NNTCATAGTGGAATTCTCGGGTGCCAAGG. RT primer sequence: GCCTTGGCACCCGAGAATTCCA; PCR primer sequences: CAAGCAGAAGACGGCATACGAGATCGTGATGTGACTGGAGTTCCTTGGCACCCGAG AATTCCA. To identify CD95L-derived small RNAs among the sequenced reads, a BLAST database was generated from each set of reads, and blastn was used to query the CD95L ORF (derived from NM_000639.2) against reads from cells infected with pLenti-CD95L and to query the CD95LMUTNP ORF sequence against reads from cells infected with CD95LMUTNP. The only reads considered further were those matching a CD95L sequence with an e-value of less than 0.05 and 100% identity across the entire length of the read. This resulted in the loss of a few reads less than 19/20 nt in length. The filtered BLAST hits were converted to a bed formatted file, describing the locations of reads relative to the relevant CD95L sequence, and the R package Sushi was used to plot the bed files and generate Figures 4B-E.
Assessing toxicity of CD95L-derived small RNAs
To determine whether guide RNAs derived from the over-expressed CD95L mRNA could evoke toxicity, the small CD95L-derived RNA reads (corresponding to different clusters shown in Figure 4C) bound to AGO from the HCT116 Drosha knock-out cells were converted to siRNAs. First, all reads less than 18 nucleotides were filtered out, as these do not efficiently incorporate into the RISC. siRNAs were designed with antisense strands identical to these CD95L-derived sequences that mapped to areas of the CD95L mRNA secondary structure (Figure 4 - figure supplement 1A) that are predicted to form duplexes. These sequences were designed as 19 nucleotide oligos with a 3’ deoxy AA. The complementary sense strand was designed with a 3’ deoxy TT and 2’-O-methylation at the first two positions to prevent its incorporation into the RISC. These oligos were ordered from IDT and annealed to form the final siRNAs. The sequences of the antisense strands (corresponding to the CD95L mRNA-derived cluster fragments) were as follows: 5’-AUUGGGCCUGGGGAUGUUU-3’ (c7/1), 5’-CCUGGGGAUGUUUCAGCUC-3’ (c7/2), 5’-CCAACUCAAGGUCCAUGCC-3’ (c11), 5’-AAACUGGGCUGUACUUUGU-3’ (c15/1), 5’-AACUGGGCUGUACUUUGUA-3’ (c15/2), 5’-CAACAACCUGCCCCUGAGC-3’ (c16/1), 5’-AACUCUAAGCGUCCCCAGG-3’ (c16/2), 5’-UCAACGUAUCUGAGCUCUC-3’ (c21), and 5’-AAUCUCAGACGUUUUUCGG-3’ (c22).
These eight siRNAs were reverse transfected into HeyA8, H460, M565, and 3LL cells using RNAiMAX transfection reagent (ThermoFisher Scientific) at 10 nM in triplicate as previously described (Murmann et al., 2018). The non-targeting (NT) and siL3 siRNAs, as described previously (Putzbach et al., 2017), were used as a negative and positive control, respectively. Cell death was quantified via ATP release 96 hours after transfection using CellTiter-Glo (Promega). The % viability was calculated in relation to the RNAiMAX-only treatment structure (Figure 4 - figure supplement 1B).
Statistical analyses
Continuous data were summarized as means and standard deviations (except for all IncuCyte experiments where standard errors are shown) and dichotomous data as proportions. Continuous data were compared using t-tests for two independent groups and one-way ANOVA for 3 or more groups. For evaluation of continuous outcomes over time, two-way ANOVA was used with one factor for the treatment conditions of primary interest and a second factor for time treated as a categorical variable to allow for non-linearity.
The effects of treatment on wild-type versus Drosha knock-out cells were statistically assessed by fitting regression models that included linear and quadratic terms for value over time, main effects for treatment and cell type, and two-and three-way interactions for treatment, cell-type and time. The three-way interaction on the polynomial terms with treatment and cell type was evaluated for statistical significance since this represents the difference in treatment effects over the course of the experiment for the varying cell types.
GSEA used in Figure 2B was performed using the GSEA v2.2.4 software from the Broad Institute (http://www.software.broadinstitute.org/gsea) 1000 permutations were used. The Sabatini gene lists were set as custom gene sets to determine enrichment of survival genes versus the nonsurvival control genes in downregulated genes from the RNA-Seq data as done previously (Putzbach et al., 2017); p-values below 0.05 were considered significantly enriched. Genes with an average normalized read expression (across both pair of duplicates) below 3 were excluded so as to only include genes that are truly expressed. The GO enrichment analysis shown in Figure 2D was performed with all genes that after alignment and normalization were found to be at least 1.5 fold downregulated with an adjusted p-value of <0.05 using the software available on www.Metascape.org and default running parameters. The other data sets used in this analysis (HeyA8 cells transfected with a toxic siRNA targeting CD95L siL3 and 293T infected with toxic shRNAs targeting CD95L shL1 and shL3 and HeyA8 cells infected with a toxic shRNA targeting CD95 shR6) were previously described (Putzbach et al., 2017).
All statistical analyses were conducted in Stata 14 or R 3.3.1.
Data availability
RNA sequencing data generated for this study is available in the GEO repository: GSE103631 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE103631, reviewer access token: etgbqyaenvirjqn) and GSE114425 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE114425, reviewer access token: edgdoaocjberbyr).
Author contributions
W.P. planned the study and performed experiments. A.H.K., Q.Q.G., A.S.Q., and A.S. performed experiments, D.M.S. provided biostatistics support, E.B. provided biocomputational support, A.A.S. performed the Ago pull down experiments, M.H. provided assistance and discussions on the mechanism of RNAi and the RISC, and M.E.P. directed the study and M.E.P. and W.P. wrote the manuscript.
Competing financial interests
The authors declare no competing financial interests.
Supplementary Videos
Video 1: CD95 k.o. HeyA8 cells (clone 11) infected with pLenti control virus.
Video 2: CD95 k.o. HeyA8 cells (clone 11) infected with pLenti-CD95Lvirus.
Video 3: HeyA8 cells infected with pLKO-shScr.
Video 4: HeyA8 cells infected with pLKO-shL3.
Acknowledgements
We are grateful to Siquan Chen for testing small CD95L-derived siRNAs. M.H. and A.A.S were supported by the Intramural Research Program of NIAMS. A.A.S. acknowledges support by the Swedish Research Council postdoctoral fellowship. This work was funded by training grant T32CA009560 (to W.P. and A.H.K.), R50CA221848 (to E.T.B.), and R35CA197450 (to M.E.P.).
Footnotes
↵4 Shared first authorship