The Drosophila Tis11 Protein and Its Effects on mRNA Expression in Flies*

Background: Insects generally express a single tristetraprolin family member, zinc finger proteins that promote mRNA decay. Results: The Drosophila protein, Tis11, can promote mRNA decay in cells, and its deficiency in flies results in accumulation of certain mRNAs. Conclusion: Tis11 deficiency in Drosophila results in increases of potential target transcripts. Significance: Tis11 can affect post-transcriptional gene expression in adult flies by regulating mRNA decay. Members of the mammalian tristetraprolin family of CCCH tandem zinc finger proteins can bind to certain AU-rich elements (AREs) in mRNAs, leading to their deadenylation and destabilization. Mammals express three or four members of this family, but Drosophila melanogaster and other insects appear to contain a single gene, Tis11. We found that recombinant Drosophila Tis11 protein could bind to ARE-containing RNA oligonucleotides with low nanomolar affinity. Remarkably, co-expression in mammalian cells with “target” RNAs demonstrated that Tis11 could promote destabilization of ARE-containing mRNAs and that this was partially dependent on a conserved C-terminal sequence resembling the mammalian NOT1 binding domain. Drosophila Tis11 promoted both deadenylation and decay of a target transcript in this heterologous cell system. We used chromosome deletion/duplication and P element insertion to produce two types of Tis11 deficiency in adult flies, both of which were viable and fertile. To address the hypothesis that Tis11 deficiency would lead to the abnormal accumulation of potential target transcripts, we analyzed gene expression in adult flies by deep mRNA sequencing. We identified 69 transcripts from 56 genes that were significantly up-regulated more than 1.5-fold in both types of Tis11-deficient flies. Ten of the up-regulated transcripts encoded probable proteases, but many other functional classes of proteins were represented. Many of the up-regulated transcripts contained potential binding sites for tristetraprolin family member proteins that were conserved in other Drosophila species. Tis11 is thus an ARE-binding, mRNA-destabilizing protein that may play a role in post-transcriptional gene expression in Drosophila and other insects.

Members of the mammalian tristetraprolin family of CCCH tandem zinc finger proteins can bind to certain AU-rich elements (AREs) in mRNAs, leading to their deadenylation and destabilization. Mammals express three or four members of this family, but Drosophila melanogaster and other insects appear to contain a single gene, Tis11. We found that recombinant Drosophila Tis11 protein could bind to ARE-containing RNA oligonucleotides with low nanomolar affinity. Remarkably, co-expression in mammalian cells with "target" RNAs demonstrated that Tis11 could promote destabilization of ARE-containing mRNAs and that this was partially dependent on a conserved C-terminal sequence resembling the mammalian NOT1 binding domain. Drosophila Tis11 promoted both deadenylation and decay of a target transcript in this heterologous cell system. We used chromosome deletion/duplication and P element insertion to produce two types of Tis11 deficiency in adult flies, both of which were viable and fertile. To address the hypothesis that Tis11 deficiency would lead to the abnormal accumulation of potential target transcripts, we analyzed gene expression in adult flies by deep mRNA sequencing. We identified 69 transcripts from 56 genes that were significantly up-regulated more than 1.5-fold in both types of Tis11-deficient flies. Ten of the up-regulated transcripts encoded probable proteases, but many other functional classes of proteins were represented. Many of the up-regulated transcripts contained potential binding sites for tristetraprolin family member proteins that were conserved in other Drosophila species. Tis11 is thus an ARE-binding, mRNA-destabilizing protein that may play a role in post-transcriptional gene expression in Drosophila and other insects.
The control of mRNA turnover rates is critical for the physiological regulation of gene expression. The rates of turnover of an individual mRNA can vary greatly under different physiological conditions and in response to a variety of stimuli. Factors contributing to mRNA lability include cis-acting sequences within mRNA, often in the 3Ј-untranslated regions (UTRs). One such category of instability-promoting sequences is the AU-rich element (ARE), 4 first described (1) as a sequence within the 3Ј-UTR of the mammalian granulocyte-macrophage colony-stimulating factor (GM-CSF) mRNA. Numerous proteins that can bind to these elements have been identified as potential mRNA-stabilizing or -destabilizing factors (2,3). One of the best characterized is tristetraprolin (TTP), an ARE-binding protein that can promote the instability of mRNAs, such as those encoding tumor necrosis factor (TNF) and GM-CSF itself as well as an increasing list of other binding targets whose identities have been established in part by studying the responses of transcripts in cells derived from TTP knock-out mice (4 -6).
TTP belongs to a small family of four related proteins in rodents, all of which have as their characteristic mRNA binding element a tandem zinc finger (TZF) domain, with identical interfinger spacing of two CX 8 CX 5 CX 3 H zinc fingers (7). The 74-amino acid TZF domain of the proteins binds to characteristic ARE sequences with low nanomolar affinity, with the preferred binding site being UUAUUUAUU (8,9); each finger binds to a UUAU half-site with virtually superimposable structures (10). In mice, disruption of the gene encoding TTP, Zfp36, demonstrated that TTP is an anti-inflammatory protein, whose absence leads to a severe, systemic inflammatory syndrome with erosive arthritis (11). Disruption of the gene encoding the second family member, Zfp36-like 1 (Zfp36l1), showed that this gene is necessary for normal allantoic fusion with the chorion and establishment of the normal feto-placental circulation (12). The third mouse family member, Zfp36l2, is important for the proliferation of early embryos past the two-cell stage (13) and is essential for hematopoiesis (14). Finally, in rodents, Zfp36l3 encodes a cytosolic, placenta-specific protein (7,15). Each protein behaves like TTP in co-transfection assays of TTP-induced mRNA instability and cell-free assays of mRNA decay (16); however, the presumed physiological targets for the other family members are largely unknown. Most vertebrates, including mammals, amphibians, and reptiles, express orthologues of the first three family members (TTP or ZFP36, ZFP36L1, and ZFP36L2). Birds appear to lack TTP (17), and amphibians and fish express one additional protein that does not have a direct mammalian orthologue (Xenopus C3H-4 (18)). We anticipate that various species will use proteins of this type in complex ways to regulate mRNA stability. However, it is impossible to predict which, if any, of the four or more physiological systems primarily regulated by the mammalian proteins will translate into non-vertebrate organisms with fewer family members.
Proteins with similar TZF domains have been identified in other eukaryotes, including plants, protists, nematodes, yeasts, and insects (19). At least one of the two such proteins expressed in Saccharomyces cerevisiae affects the stability of a collection of transcripts involved in regulating iron metabolism, apparently also by binding to ARE elements that are similar to their mammalian counterparts (20,21). In Schizosaccharomyces pombe, the single family member, Zfs1p, also binds to and promotes the decay of ARE-containing mRNAs (22,23), but in this case, the major phenotypic effect of the protein is to prevent abnormal cell-cell interactions, and iron metabolism appears to be unaffected. This evidence strongly suggests that at least some of the functions of proteins from this family have been conserved in eukaryotes, such as the ability to bind to and destabilize ARE-containing mRNAs, but that these proteins can be used for different physiological purposes in different species.
Most insect genomes appear to contain only a single gene expressing proteins of this type. In Drosophila melanogaster, for example, the protein product of the Tis11 locus contains a TZF domain with identical critical residues and spacing to the mammalian members of the TTP family (24). The effects of the encoded protein, Tis11, have been studied recently in Drosophila S2 cells, with an emphasis on the possible relevance to Drosophila innate immunity, based on the analogy with TTP itself (25)(26)(27)(28)(29)(30). However, very little is known about the physiological role of Tis11 in intact flies. Because of the apparent expression of only a single such gene in most if not all insect species, the genetic tractability of Drosophila, and the economic and medical importance of many other insects, we have begun to explore the physiological role of the Tis11 protein in flies. We show here that Tis11 resembles the mammalian proteins in many respects, in that it exhibits high affinity ARE-binding through its TZF domain and has mRNA-deadenylating and -destabilizing properties that are in part due to the presence of a C-terminal potential Not1 binding domain. Two different genetic models of loss of function in adult flies suggest that, as expected, it can regulate post-transcriptional gene expression in Drosophila. Strikingly, the transcripts identified in adult flies as potential Tis11 targets were almost entirely different from those identified as targets in S2 cells, suggesting that both approaches may be complementary in elucidating the physiological role of this protein in insects.

Expression Plasmids
CMV.dTis11-The template used to amplify the Tis11 protein coding fragment was an expressed sequence tag clone purchased from the IMAGE consortium (GenBank TM accession number AI512677). This clone contained the entire predicted D. melanogaster Tis11 variant A protein coding sequence (GenBank TM accession number NP_511141.2). The PCR primers used were 5Ј-ggtaccACAAAATGTCTGCTGATATT-CTG-3Ј (5Ј) and 5Ј-ctcgagTTAGAGTCCCAAATTGGACTGC-TGC-3Ј (3Ј), where the capital letters correspond to the Tis11 sequences and the lowercase letters correspond to the recognition sites for Asp718 (the 5Ј primer) and XhoI (the 3Ј primer). The PCR product was digested with Asp718 and XhoI and ligated into the vector CMV.BGH3Ј/pBSϩ (31).
CMV.dTis11.HA/His-In this construct, the hemaglutinin-His 6 (HA/His 6 ) epitope tags were linked to the last amino acid of the Tis11 protein by the polymerase chain reaction primeroverlapping mutagenesis technique (32), and a BamHI site was created between the Tis11 C terminus and the epitope tag. The fusion cDNA was then ligated into the Asp718 and XhoI sites of the vector CMV.BGH3Ј/pBSϩ. Key cysteine and histidine residues in the Tis11 TZF domain were mutated to other amino acids to create potential non-RNA-binding mutants, using the PCR primer-overlapping mutagenesis technique. The primer representing the upper strand sequence used in the mutation of cysteine 150 to a serine (C150S) was 5Ј-GCCGGAGAATCCA-AGTACGGCG-3Ј, in which the bases encoding serine are underlined. To mutate histidine 198 to a glutamine (H198Q), the primer representing the upper strand sequence was 5Ј-CACT-TTGTTCAGAATGCGGAC-3Ј, with the underlined bases encoding glutamine.
CMV.dTis11.GFP-An Asp718-BamHI fragment from the construct CMV.dTis.HA/His, which contained the entire protein coding region except the stop codon (with or without the mutated key cysteine and histidine residues), was ligated into the vector CMV.EGFP.BGH3Ј/BSϩ for the expression of the Tis11.GFP fusion protein. The vector CMV.EGFP.BGH3Ј/BSϩ was made by ligating a BamHI-NotI fragment from pEGFP-N1 (Clontech), containing the protein coding region for EGFP, into the BamHI and NotI sites of CMV.BGH3Ј/pBSϩ. CMV.HA.dTis11 and Its C-terminal Truncation Mutants-These constructs were used to study the involvement of the conserved C-terminal domain in the ability of Tis11 to destabilize ARE-containing mRNAs. The conserved C-terminal motif of nine amino acids, RLP(I/V)FNR(I/L)S, recently implicated in the binding of human TTP to NOT1 (33), is located at 414 -422 of the Tis11 isoform A (GenBank TM accession number NP_ 511141.2). A 5Ј primer containing the Asp718 restriction enzyme site and encoding the Tis11 isoform A residues 1-7, 5Ј-ggtaccATGTCTGCTGATATTCTGCAG-3Ј, was used to create the following expression constructs. For full-length CMV.HA.Tis11, the 3Ј primer used for plasmid CMV.dTis11 was used. For the truncation plasmid CMV.HA.Tis11⌬414 -422, a pair of complementary primers were used that encoded the sequences for Tis11 residues, 408 QQEDTP-deletion-SGVEAY 428 , in combination with the 5Ј and 3Ј primers described for the CMV.HA.dTis11 plasmid, using the PCR primer-overlapping technique (32). The upper strand sequence was 5Ј-CAGCAG-GAGGATACGCCC,TCCGGTGTGGAGGCCTAC-3Ј, with the comma denoting the site of the deletion.
For the truncation plasmid CMV.HA.Tis11⌬414 -436, a 3Ј PCR primer containing the sequences for residues 406 -413, the stop codon, and the XhoI restriction enzyme was used. The PCR products were digested with Asp718 and XhoI and ligated into the vector CMV.HA.BGH3Ј/pBSϩ, 5 in frame with the HA tag at the Tis11 N terminus. The RNA-degrading activities of Tis11 expressed from these constructs were assessed by cotransfecting them with the Mlp-TNF3Ј target plasmid, in which a portion of the mouse Mlp promoter and the protein coding region of the mouse Mlp mRNA is followed by a 3Ј-UTR consisting of a portion of the mouse Tnf mRNA 3Ј-UTR, as described in detail previously (15). The mouse Mlp promoter is very sensitive to actinomycin D, in contrast to the CMV promoters used for the other expression plasmids used in this study.
For bacterial expression, the cDNA clone CMV.dTis11 was used as a template for PCR; the amplified fragment containing the entire open reading frame was cloned into the expression vector pET21a in frame with a His 6 tag on the 3Ј-end by standard procedures (Abgent, San Diego, CA). The resulting expression plasmid was used for the expression in and purification of the Tis11-His 6 fusion protein from E. coli using nickel affinity columns (Abgent). This purified protein was used either in gel shift RNA binding studies, using a mouse Tnf mRNA-based cDNA probe as described (31), or in quantitative fluorescence anisotropy binding studies as described below. In addition, a total of 825 g of purified protein per rabbit was used to produce polyclonal antisera according to standard procedures (Covance, Princeton, NJ).

RNA-Protein Binding Affinity
Quantitative measurements of RNA-protein binding affinity were performed using a fluorescence anisotropy-based assay essentially as described (34). Briefly, a fluorescein-tagged RNA substrate (Fl-ARE, with sequence Fl-AUUUAUUUAUUUA; 0.2 nM) containing a single high affinity binding site for the human TTP TZF domain (35) was incubated at 25°C with varying concentrations of recombinant Tis11 in a final volume of 100 l containing 10 mM Tris⅐HCl (pH 8.0), 100 mM KCl, 2 mM dithiothreitol, 0.1 g/l bovine serum albumin, 5 M ZnCl 2 , and 0.2 g/l heparin. Fluorescence anisotropy was measured with a Beacon 2000 fluorescence polarization system (Panvera, Madison, WI) using fluorescein excitation ( ex ϭ 490 nm) and emission ( em ϭ 535 nm) filters. Protein binding increases the measured anisotropy of the fluorescein-conjugated RNA, based on restriction of segmental motion and rotational correlation time (36). Tis11 binding did not influence the fluorescence quantum yield of the Fl-ARE substrate (data not shown), permitting the bimolecular association constant (K a ) to be resolved from binding reactions across a titration of protein ([Tis11]) using Equation 1 (34), where A t represents the total measured anisotropy, and A R and A PR represent the intrinsic anisotropy values of the free and protein-associated Fl-ARE substrates, respectively. A R was measured as the anisotropy of the Fl-ARE substrate in the absence of protein (n Ն 3), whereas all other constants were solved by non-linear least squares regression of A t versus [Tis11] using PRISM version 3.0 (GraphPad, San Diego, CA). Complex dissociation constants (K d ) were resolved using K d ϭ 1/K a .

Structural Modeling Studies
Tis11-RNA Complex-An initial model of the TZF domain of D. melanogaster Tis11 (amino acids 133-202 from GenBank TM accession number AAF48194.2) was created using the RNAbound structure of the TZF domain of ZFP36L2 (TIS11d) (Protein Data Bank entry 1RGO (10)). Necessary mutations were introduced into the sequence of human ZFP36L2 using Coot (37) to generate a model of the Tis11 TZF domain in its RNAbound form (with two CCCH coordinated zinc ions) and energy-minimized using the program Amber.12. The RNAbound Tis11 TZF complex was solvated in a box of water (about 9500 water molecules). Prior to equilibration, the solvated system was subjected to 1) 100-ps belly dynamics runs with fixed peptide, 2) minimization, 3) low temperature constant pressure dynamics at fixed protein to assure a reasonable starting density, 4) minimization, 5) stepwise heating molecular dynamics at constant volume, and 6) constant volume molecular dynamics for 1 ns. A final unconstrained molecular dynamics trajectory was calculated at 300 K under constant volume (15 ns, time step 1 fs) using PMEMD (Amber.12) to accommodate long range interactions. The force field parameters were taken from the FF03 force field (38) for the protein and the PARMBSC0 force field (39) for the RNA. The charge on zinc ions was ϩ2, whereas cysteine coordinated with zinc carried a Ϫ1 charge. His coordinated to zinc remained neutral. The van der Waals parameters of zinc were adjusted to yield Zn-S and Zn-N distances between 2.25 and 2.50 Å during the simulations. No specific constraints were applied to maintain the zinc coordination. Images of molecular models were created using VMD.1.9.1 (40).
Human TTP (or Tis11) Bound to Human NOT1 or Drosophila Not1-Using the x-ray crystal structure (Protein Data Bank entry 4J8S) of human TTP bound to human NOT1 (residues 820 -999) (33), an initial model for the conserved Tis11 C-terminal domain bound to the TTP binding domain of human NOT1 was created. Necessary mutations were carried out using the program Coot (37). Using the coordinates of human TTP bound to human NOT1 as a template, Coot was employed to construct an initial model of Tis11 bound to Drosophila Not1. The new models of Tis11 bound to human NOT1, Tis11 bound to Drosophila Not1, and the original model of human TTP bound to human NOT1 (33) were then solvated separately in a box of water, and the solvated systems were subjected to the protocol described above, followed by a long molecular dynamics simulation at T ϭ 300 K using the "PMEMD" module of the program Amber.12. The final snapshots of the bound solution structure (after 25 ns) were selected for comparison.

Cell Transfections
For mammalian cell transfection experiments, human embryonic kidney 293 cells (referred to as 293 cells) were transfected with the plasmids described above that expressed either the epitope-tagged normal Tis11 protein, its mutant derivatives, or the untagged protein and its C-terminal truncations, driven by the CMV promoter. In one case, 293 cells were cotransfected with the plasmids expressing the Tis11 fusion proteins along with a somewhat shortened version of the mouse Tnf mRNA, which contains several potential TTP binding sites but excludes other AU-rich sequences, as described in detail previously (31). This represents one assay for the mRNAdestabilizing activities of TTP family member proteins. A plasmid encoding human TTP was transfected in parallel experiments as a positive control. In this case, the assay readout consisted of Northern blots probed with radiolabeled probes for TNF, as described (31); in addition, expression levels of TTP and Tis11 mRNA were determined by Northern blotting, using total cellular RNA isolated from the transfected cells as described (31).
In another series of transfection experiments, CMV-driven Tis11 and its C-terminal truncations were co-transfected with constructs in which an Mlp-Tnf 3Ј-UTR fusion transcript was driven by the mouse Mlp promoter, as described (15). The ability of Tis11 constructs to promote decay of the Mlp-Tnf3Ј fusion mRNA was assessed by Northern blotting with an Mlp cDNA probe, as described (15).
In other transfection experiments, plasmids encoding either epitope-tagged human TTP or Tis11 were transfected into 293 cells, and the soluble cellular supernatants were used in RNA gel shift assays as described (31). This assay used a radiolabeled probe based on the core ARE sequence of the mouse Tnf 3Ј-UTR (31).

Cell-free Deadenylation Assays
In these assays, performed essentially as described (41), the RNA (8 g/tube) was incubated in 10 l of 50 mM KCl without (Ϫ) or with (ϩ) 0.6 g of dT 12-18 at 95°C for 5 min, followed by another 30 min at 25°C. A 15-l aliquot of buffer was then added to each tube, without (Ϫ) or with (ϩ) 1 unit of RNase H (New England Biolabs) and incubated for 45 min at 37°C. The final incubation buffer contained 30 mM Tris-HCl (pH 8.3), 65 mM KCl, 4.8 mM MgCl 2 , 6 mM DTT, and bovine serum albumin (0.5 g/l).

RNA Decay Assays
Each plate of 293 cells was co-transfected with 3 g of the Mlp-Tnf3Ј plasmid and 2 g of vector (BSϩ) DNA. In addition, 5 ng of an EGFP construct was added along with 5 ng of CMV/ BSϩ vector DNA (control group), 10 ng of CMV.DTis11.HA/ His 6 (WT Tis11 group), or a construct encoding the Tis11 mutant H198Q (mutant group). Twenty-four h after removing the transfection mixture, RNA was isolated from three plates of cells from each group, and actinomycin D (5 g/ml) was added to the rest of the plates. At each indicated incubation period, three plates of cells from each group were used for RNA isolation, and this was used for Northern blotting with a mouse Mlp cDNA probe. These were quantitated using a PhosphorImager and ImageQuant software (GE Healthcare). The Phosphor-Imager volume of the transfected-expressed Mlp-Tnf3Ј mRNA at each gel lane was normalized with the volume of the endogenous MLP mRNA in the same lane. Similar transfection experiments were performed five times.

Drosophila Stocks
The P element insertion stock w* P[EP]Tis11 G147 (referred to below as G147) was purchased from GenExel (Daejeon, South Korea); this P element was described by GenExel as within the first exon of Tis11. Df(1)Tis11-IE35 (hereafter known as IE35) was identified in an imprecise excision screen performed by mobilizing the P element in G147 with the ⌬2-3 transposase, followed by screening for lethals over Df(1)N105, obtained from the Bloomington Drosophila Stock Center. In this screening, with the help of a mus309 allele mutation, large deletions were created. Dp(1;Y)BSC5 (hereafter known as BSC5) is a Y chromosome bearing an X chromosome duplication from 11B9 to 11D5 and a second duplication bearing y ϩ and was obtained from the Bloomington Drosophila Stock Center.
We also obtained the P element insertion stock w* P[EP]-Tis11 G1183 /FM7c (referred to below as G1183) from GenExel. This P element was also described as within the first exon of the Tis11 gene. G1183 was reported by GenExel to be homozygous lethal, and the orientation of the UAS sequence within the EP element (42) should drive transcription in the opposite direction relative to the Tis11 transcript when exposed to Gal4 drivers. G147 was described as viable and is expected to drive transcription in the 5Ј-3Ј direction relative to Tis11. The P element insertion sites and orientation were confirmed by direct sequencing from both directions and are as follows. For G147, the insertion site is between bp 12553436 and 12553437 of the X chromosome sequence NC_004354.3 or between bases 42 and 43 of the Tis11 mRNA sequence NM_078586.3. For G1183, the P insertion site was between bp 12553497 and 12553498 of the X chromosome sequence NC_004354.3 or between bases 103 and 104 of the Tis11 mRNA sequence NM_078586.3.
The original lethal G1183 chromosome became viable during maintenance in our facility and was termed G1183v. We therefore tried to eliminate possible mutation(s) other than the P element in the original chromosome by allowing meiotic recombination with a chromosome marked with w 1118 . The resulting viable recombinant chromosome was termed "G1183rec." This chromosome still contained the w ϩ P element but was viable, suggesting that the lethality of the original G1183 chromosome resulted from one or more additional mutations outside the Tis11 locus. These flies were then backcrossed to y w 67c23 flies for 10 generations to form the G1183 and control flies used for mRNASeq in this paper.

Next Generation Sequencing
Two mRNASeq experiments were performed for this analysis, carried out by the National Institutes of Health Intramural Sequencing Center. In one experiment, 1-5-day-old w* IE35/ BSC5 male flies (referred to below as "null" flies) and w 1118 / BSC5 males (referred to as WT or control flies) were collected into separate pools of 40 flies and frozen until use. RNA was isolated from flies using the GE Healthcare RNA isolation kit. The RNA was quantitated using a Qubit fluorometer (Invitrogen). Ten g of total RNA was transcribed to cDNA and subjected to Illumina deep sequencing analysis using 50-bp paired end reads, generated from an Illumina Genome Analyzer II. There were four separate pools of null and control flies in this experiment.
In a second experiment, 1-5-day-old male y w 67c23 G1183 flies from the 10-generation backcross and their y w 67c23 controls were used to obtain RNA in the same way and subjected to deep sequencing using 100-bp paired end reads generated from an Illumina HiSeq 2000 system. In this experiment, there were also four separate pools of G1183 and control flies.
Raw sequence reads were mapped to the cDNA sequences from the Drosophila genome using BWA (44), with the default setting of its paired end read mapping protocol. This was done separately for each replicate. The reference cDNA sequences were based on the Drosophila genome annotation (version r5.49) from FlyBase. After read mapping, we used the improved version of EpiCenter (45) and the statistical computing language R to perform statistical analyses to identify differentially expressed genes. Specifically, we first filtered out all reads that were not mapped by pairing to the reference cDNA sequences and then counted reads mapping to each cDNA sequence. To quantitate individual transcripts for a "transcript level" analysis, we counted all reads mapped to the entire cDNA sequence of the transcript. For a "gene level" analysis, we counted only reads that were mapped to the common exons shared among all transcript variants from the gene. Read counts of mRNA in each replicate sample were then normalized by the average number of total mapped reads per replicate in both the experimental and control samples. Any transcripts with normalized read counts of less than 100 in both the experimental and control samples were removed from further analysis. For the remaining transcripts, we compared mean expression levels of transcripts that differed significantly between IE35/BSC5 and w1118/ BSC5 flies, using unpaired, one-or two-tailed Student's t tests. Similar comparisons were performed between the mean expression values from the G1183 flies and their respective controls. In both cases, we also used p values adjusted for multiple testing based on EpiCenter's Max-P statistic, which takes into account the total number of tests (45).

NanoString Assays
For confirmation assays, we used the NanoString nCounter method (46) to analyze a subset of the transcripts that were increased in both types of mutant flies as well as two Tis11 transcript variants, Tis11-RA and Tis11-RC, and the CecA1 (Cecropin A1) mRNA. This method uses solution phase hybridization of RNA to pairs of specific oligonucleotide probes and provides digital quantitation of mRNA levels without amplification. We used the same RNA samples that were used in the mRNASeq analyses, except that only three of the null samples were available.

RESULTS
The D. melanogaster Tis11 Protein-According to current models in FlyBase, the mRNA products of Tis11 are labeled Tis11-RA, -RC, and -RD. RD is truncated at the extreme 5Ј-end, but this does not affect the predicted protein sequence. RC has a much longer 3Ј-UTR than the other two forms but otherwise is identical to Tis11-RA. All three encode the same 436-amino acid protein, which we will refer to as Tis11. An expressed sequence tag clone corresponding to the original A variant was therefore used to make the clones for the transfection and expression studies described here. Although FlyBase and GenBank TM differ somewhat on the isoform nomenclature, the A variant is the same in both databases and is represented by GenBank TM RefSeq numbers NM_078586.4 and NP_511141.2 for the mRNA and protein, respectively.
An alignment of the Drosophila Tis11 protein with the shortest insect orthologues we have identified to date, from Bombyx mori and Tribolium castaneum (253 and 250 amino acids, respectively), is shown in Fig. 1A. There is near sequence identity in the TZF domain among the proteins from the three species. There is also a highly conserved region near the extreme C terminus (green). The conserved C-terminal region in mammalian ZFP36L1 and ZFP36L2 is known to contain an active nuclear export sequence (7,47). Recent studies in Drosophila FIGURE 1. Alignment of Drosophila Tis11 with related insect proteins and human TTP family members. A, ClustalW2 alignment of the Drosophila Tis11 protein (GenBank TM accession number NP_511141.2) with the corresponding proteins assembled from Bombyx mori (XP_004928274) and Tribolium castaneum (XP_968440). These two insects were chosen because their predicted protein sequences are the shortest we have found in insects to date. Asterisks indicate amino acid identity at that site; single and double dots indicate weak and strong amino acid similarities, respectively. The residues within the tandem zinc finger domains that are important for zinc coordination are in red; the conserved C-terminal domains discussed under "Materials and Methods" in green. The yellow shading indicates the nuclear export sequence identified in Ref. 48 have confirmed that Tis11 is a nucleo-cytoplasmic shuttling protein and have localized the active nuclear export sequence to amino acids 101-113 (Fig. 1A, yellow) (48). This sequence is not particularly well conserved in the other insects in the alignment shown, but there are neighboring regions of alternating hydrophobic residues that may turn out to be nuclear export sequences. The conserved C-terminal region in human TTP has recently been shown to be responsible for binding to the NOT1 scaffolding protein and appears to be involved in recruiting deadenylases to the protein-RNA complex (33). Outside of these two domains, there was remarkably little sequence conservation between the protein from D. melanogaster and those from the other two insects.
We also aligned the Drosophila Tis11 protein with the three known human TTP family members: TTP (ZFP36), ZFP36L1, and ZFP36L2 (Fig. 1B). A BLAST comparison of the full-length proteins suggests that the Tis11 protein is most closely related to the human ZFP36L1 protein (2EϪ43), and less closely related to ZFP36L2 (1EϪ39) and TTP (1EϪ26). As with the other insects, the predominant regions of conservation are the TZF domain and the extreme C-terminal domain (green). The experimentally demonstrated nuclear export sequences in these proteins are labeled in yellow; note the close overlap between these sequences and the probable NOT1 binding sequences in human ZFP36L1 and ZFP36L2.
For the purposes of this study, in view of the experiments involving C-terminal deletions in Tis11 described below, it is important to note that the conserved C-terminal sequence in Drosophila does not appear to contain the nuclear export sequence in this species. In experiments not shown here, we have confirmed the findings by Twyffels et al. (48) that Drosophila Tis11 is a nucleo-cytoplasmic shuttling protein and that the C-terminal conserved domain does not function as a nuclear export sequence.
Expression of Tis11 in Human Embryonic Kidney 293 Cells-The human embryonic kidney 293 cells (referred to as 293 cells) used in our laboratory are useful for expression studies in this context because they do not express detectable levels of any of the three human TTP family member mRNAs. When the Tis11 protein was expressed in these cells either with ( Fig. 2A, lane 2) or without a His 6 epitope tag ( Fig. 2A, lane 3) and blotted with a Tis11 antibody, both proteins exhibited an apparent M r of ϳ55,000 -70,000 on SDS gels ( Fig. 2A). The HA/His 6 epitopetagged protein migrated at a slightly higher apparent M r and was approximately as well expressed as the untagged native protein in this transfection system ( Fig. 2A). No corresponding band was seen in cells transfected with the control vector (BSϩ, lane 1) or on a similar blot incubated with preimmune serum under the same conditions (data not shown). The difference between these values and the predicted size of the untagged protein (M r 47,166) is probably due to anomalous migration and presumed phosphorylation, as has been seen with the mammalian TTP family members. The possibility of phosphorylation is supported by both the broad band or multiple bands of the immunoreactive protein seen on one-dimensional SDS gels ( Fig. 2A) and the multiple isoelectric species seen on two dimensional electrophoresis (Fig. 2B). This pattern of multiple bands spanning considerable electrophoretic distance has been observed with the mammalian proteins and is thought to reflect multisite phosphorylation (49,50). The observed two-dimensional gel migration resulted in an approximate isoelectric point of 7.4, which matched well to the predicted pI of 7.4 -8.1. Importantly, there was very little apparent proteolysis of the full-length fly protein in this mammalian cell expression system (Fig. 2, A and B), at least that reacted with the polyclonal antiserum.
The Tis11 protein expressed in 293 cells was then tested for its ability to bind to AU-rich elements (AREs) in mRNAs. As shown in Fig. 2C, both the GFP and the HA/His 6 fusion proteins with Tis11 readily bound to RNA probes carrying the ARE sequence derived from the mammalian TTP binding sites in TNF (31) (Fig. 2C, lanes 4 and 7). Human TTP was used as a positive control (Fig. 2C, lane 2). To confirm that this binding was due to the presence of the key zinc coordinating residues in the TZF domain, we mutated two of the conserved cysteine or histidine residues within both human TTP and the Tis11 fusion proteins (Fig. 2C, lanes 3, 5, 6, and 8). Although these mutant proteins were expressed in mammalian cells approximately as well as the wild-type fusion proteins, as determined by Western blotting (Fig. 2, D and E), they each exhibited essentially no binding to the ARE-containing RNA probes under these conditions (Fig. 2C, lanes 3, 5, 6, and 8).
Expression of Recombinant Tis11 and RNA Binding-We also expressed the full-length His 6 -tagged protein in E. coli and purified the resulting expressed protein (Fig. 2F, inset); this was used in a quantitative binding assay to RNA oligonucleotides using fluorescence anisotropy (Fig. 2F). In these experiments, binding of the full-length recombinant fusion protein to a 5Ј-fluorescein-labeled RNA oligonucleotide based on the mammalian TNF TTP binding site revealed a K d of 2.3 Ϯ 0.1 nM at 25°C (n ϭ 2), almost identical to that seen with synthetic peptides derived from the human TTP TZF sequence (8,35).
When the structure of the Tis11 TZF domain bound to the canonical TTP binding motif UUAUUUAUU was modeled after the solution structure of the human TTP relative ZFP36L2 (TIS11D) (10), a nearly identical predicted structure was produced ( Fig. 2G). In particular, the aromatic residues Tyr 152 and Tyr 190 (yellow), and Phe 158 and Phe 196 (green), which are thought to be important for stacking interactions between the aromatic rings and the RNA bases, were identical to those found in the mammalian family members.
Effect of Tis11 on mRNA Stability in HEK 293 Cells-These data demonstrated that full-length Tis11 could bind to RNA ARE sequences with high affinity. However, to address whether Tis11 is an mRNA-destabilizing protein, we first expressed it in 293 cells in the presence of "target" RNAs, containing portions of the mammalian TNF 3Ј-UTR ARE, that have been validated as sensitive targets of mammalian TTP and its family members (31). We expressed human TTP using the same vector as a positive control. Under these conditions, transfection of low amounts of DNA (5 ng of DNA/plate of cells) of both human TTP (Fig. 3A, lane 2) Fig. 3A), which then accumulated to an even greater extent at higher levels of transfected Tis11 DNA. This paradoxical increase in the apparently deadenylated target transcript at higher concentrations of transfected DNA has previously been seen with other TTP family members in this assay (31,51). Surprisingly, transfection of the same low amount of the human TTP DNA plasmid (Fig. 3A, lane 2) caused approximately the same extent of transcript depletion as Tis11 (Fig. 3A, lane 3), not only indicating that the Drosophila protein was effective at promoting transcript decay in this experimental system but suggesting that it was active at roughly the same concentrations as human TTP. Fig. 3, B and C, demonstrates the expression of both TTP and Tis11 mRNAs in this experiment.
To confirm that the lower band seen after TTP or Tis11 expression was the deadenylated form of the target transcript, we performed RNase H/oligo(dT) incubations with the RNA samples shown in Fig. 3A, as described previously (41). Fig. 3D shows a Northern blot, using a TNF cDNA probe, of the RNA samples shown in Fig. 3A, both before and after treatment with RNase H. As shown in the pair of lanes labeled 1, using an RNA sample from the original lane 1 in Fig. 3A, RNase/  3D), supporting their identification as deadenylated forms of the TNF transcript.
These data established that Tis11 is an ARE-binding and mRNA-destabilizing protein that can behave the same way as the mammalian TTP family proteins to decrease the levels of, and apparently deadenylate, target mRNAs in this co-transfection assay in a heterologous mammalian cell system.
These results suggested that conserved sequence domains between the fly and mammalian proteins might be involved in, for example, recruiting deadenylase complexes to the Tis11 bound to mRNA. Because the only apparently conserved sequences outside of the TZF domain were in the extreme C terminus (see Fig. 1B), we deleted this conserved domain in the untagged Drosophila protein and examined the effects of these deletions on Tis11 activity in the 293 cell co-transfection system. Based on similar analyses done with the human TTP protein (33), we hypothesized that the conserved C-terminal motif would be important for the mRNA-degrading activity of Tis11, possibly by recruiting a similar Not1 complex. The sequence of the C-terminal end of the Tis11 protein, from residue 414 to the end, is shown in the top right-hand corner of Fig. 3; underlined in boldface type are the residues of the putative Not1 binding sequence discussed below.
As shown in Fig. 3E, using a mouse Mlp mRNA protein coding domain fused to a portion of the mouse Tnf mRNA as a target transcript (33), the N-terminal HA-tagged Drosophila Tis11 protein was effective in decreasing the steady-state levels of the target transcript (Fig. 3E, compare lanes 1 and 2). On average, this decrease was ϳ53.5 Ϯ 10% under these conditions (p Ͻ 0.0001, using a two-tailed t test) (Fig. 3G). When the last 23 amino acids of the C terminus (Arg 414 -Leu 436 ) were removed or when the nine amino acids (Arg 414 -Ser 422 ) that comprise the conserved putative Not1 binding domain were removed by an internal deletion, the protein was less effective at promoting mRNA decay than the WT protein (target transcript decreased by 30.2 Ϯ 11.4 and 26.5 Ϯ 7.5%, respectively) (Fig. 3, E (lanes 3  and 4) and G). When the effectiveness in promoting mRNA decay was compared, both of these deletion mutants were significantly less effective than the WT protein (one-way analysis of variance, Tukey's multiple comparison test; Fig. 3G). However, there was no significant difference between these two mutants, suggesting that the 14 amino acids (Ser 423 -Leu 436 ) immediately C-terminal of the conserved domain were not critical for the induced mRNA decay. As expected, when the nonbinding zinc finger mutant H198Q protein was co-expressed with the target transcript, there was no effect on steady state levels of the target mRNA compared with EGFP alone (no Tis11) (Fig. 3E, compare lanes 1 and 5; also see Fig. 3G). The mutant proteins were expressed in these experiments at comparable levels (Fig. 3F).
To examine the effect of WT Tis11 on mRNA decay, we performed separate co-transfection experiments and examined the time course of decay of the Mlp-Tnf3Ј fusion mRNA decay in the presence of WT Tis11 or one of its non-binding mutants (Fig. 3H). This construct uses the mouse Mlp promoter to drive the expression of a fusion transcript comprising mouse Mlp mRNA and a portion of the mouse Tnf mRNA (33); this pro-moter is inhibited by actinomycin D, in contrast to the CMV promoter used elsewhere. Each plate of 293 cells was co-transfected with 3 g of the Mlp-Tnf3Ј target plasmid and 2 g of vector (BSϩ) DNA and, as noted in the figure, 5 ng of a EGFP construct with 5 ng of CMV/BSϩ vector DNA (GFP), 10 ng of CMV.DTis11.HA/His6 (TIS11 WT), or 10 ng of the TIS11 construct containing the H198Q mutation (H198Q). Twenty-four hours after removing the transfection mixture, actinomycin D (5 g/ml) was added to each plate, and samples were prepared at the indicated times for RNA preparation. Each sample represents RNA from three plates of cells. A representative Northern blot from these experiments was probed with a mouse Mlp cDNA probe (Fig. 3H); the band labeled endo refers to endogenous human MLP mRNA, and the band labeled Mlp-Tnf3Ј indicates the fusion target mRNA. Under these conditions, the transfected Mlp-TNF3Ј mRNA was reasonably stable for 4 h after the actinomycin D addition in the GFP-transfected cells (Fig. 3H, left). When WT Tis11 was expressed, there was a marked decrease in the steady state levels of the transfectedexpressed Mlp-Tnf3Ј mRNA as well as an apparent rapid disappearance of this transcript with time, with perhaps slower disappearance of the lower, deadenylated form of the transcript (Fig. 3H, middle). In contrast, the co-transfection of the H198Q mutant Tis11 protein did not affect either the steady state levels of target mRNA before actinomycin D treatment or the apparent decay of the target mRNA after actinomycin D (Fig. 3H,  right). Endogenous levels of human MLP mRNA did not change markedly during these manipulations and served as an internal loading control. We also performed RNase H-oligo(dT) assays on the samples from this experiment taken from the cells before the actinomycin D treatment and found migration patterns of the polyadenylated and deadenylated mRNA species that were similar to those shown in Fig. 3D (data not shown).
Northern blotting data from five such experiments were quantitated using a PhosphorImager and were averaged in Fig.  3I, with the steady state levels before actinomycin D normalized to 1.0 at 0 h. Means Ϯ S.D. were plotted in this figure. When either GFP or the H198Q mutant of Tis11 was co-transfected, the decay lines were superimposable and decreased to ϳ70% of time 0 after 4 h. In contrast, the target transcript decayed rapidly in the Tis11-transfected cells, reaching 50% of the original level after 50 min. These data support the concept that Tis11 promoted the decay of this ARE-containing transcript in a manner that is completely dependent on the integrity of the TZF domain.
Structural Models of the Drosophila Tis11 Putative Not1 Binding Domain-To explore further the feasibility of the Tis11 C-terminal domain as a Not1 binding domain, we performed simulation solution modeling of the proposed association between the Tis11 C-terminal peptide and the putative Drosophila Not1 Tis11 binding domain, based on the original coordinates of Fabian et al. (33). The C-terminal sequences from human TTP and Drosophila Tis11 are aligned in Fig. 4A, with the demonstrated human NOT1 and putative Drosophila Not1 binding domains in boldface type; the symbols are described in the legend to Fig. 1. Fig. 4B shows the proposed structures of the binding complex formed by either the human TTP C-terminal peptide (red) or the Drosophila Tis11 C-terminal peptide Drosophila Tis11 DECEMBER 19, 2014 • VOLUME 289 • NUMBER 51

JOURNAL OF BIOLOGICAL CHEMISTRY 35051
(green) to the TTP binding domain of human NOT1 (silver) and Drosophila Not1 (blue), demonstrating that the proposed Drosophila complex is structurally very similar to the complex formed between the two human proteins (33). In fact, essentially all of the key residues in the human protein complex are present in the putative Drosophila protein complex (Fig. 4B). In Fig. 4C, we also modeled the proposed interaction between the Drosophila Tis11 C-terminal peptide shown in A (green) with the human NOT1 TTP binding site (silver). The potentially interacting side chains of NOT1 (gray) and the Tis11 peptide (colors, with green labels) are shown as sticks. Hydrophobic interactions, salt bridges, and hydrogen bonds, shown in magenta, presumably contribute to the predicted stability of the interface (Fig. 4C).
These data suggest that the conserved C-terminal domain in Drosophila is likely to serve as the binding site for either the human or Drosophila Not1 protein, which in turn can serve as the scaffolding for the entire CCR4-NOT1 deadenylating complex. As with the human TTP protein (33), deletion of the presumed Not1 binding domain of Tis11 does not totally abrogate its mRNA-degrading activity, suggesting that other binding sites for other mRNA-degrading activities may exist in both proteins. Importantly, the deletions that impaired mRNA deadenylating activity had no effect on nucleoplasmic shuttling (data not shown; also see above), excluding this as a potential artifact in the interpretation of these experiments.

Loss of Function Experiments in Adult Male
Flies-Tis11 is on the Drosophila X chromosome, and Fig. 5A shows a diagrammatic representation of this locus and neighboring loci (Ck1␣ and Smr), with the centromere to the right. The two EP elements discussed in this paper are shown at the top, with the vertical dotted lines indicating their position in the first exon of Tis11 and ultimately in the 5Ј-UTR of the Tis11 mRNA, as well as their directions of transcription. The approximate locations of the IE35 imprecise excision and the BSC5 duplication are also indicated. Transcription of Tis11 occurs from left to right in this figure. Note the broken lines, indicating the removal of 15 kb from the large second intron of Tis11.
To begin to explore the possible function of Tis11 in intact flies, we conducted an imprecise excision screen by mobilizing the P element in G147 flies, and chromosomes were screened over Df(1)N105. In this screen, lethality was used to identify imprecise excision, because, given the lethality seen in the initial G1183 stock obtained from GenExel, we assumed that a Tis11 deficiency would have a lethal phenotype. All 45 of the lethal chromosomes recovered were also deficient for one of the neighboring genes, Ck1␣ or Smr. One of the imprecise excision events, Df(1)Tis11-IE35, was balanced over FM7i, Act5C-GFP. The IE35 hemizygous males died at the first or second larval instar. To map the extent of the IE35 deficiency, first instar larvae were used for the purification of genomic DNA and analyzed by PCR. This analysis confirmed that IE35 was deficient in both Tis11 and Smr. Real-time RT-PCR of IE35 males carrying Dp(1;Y)BSC5, a Y chromosome with a duplication of a portion of the X chromosome containing Smr, demonstrated the absence of Tis11 transcripts, whereas CK1a and Smr transcript levels were not affected (Fig. 5B). Further PCR analyses showed that IE35 is a deletion of ϳ200 kb that extends from the first exon of Tis11, where the original P element was inserted, proximally to CG4661. Analysis of Dp(1;Y)BSC5, using genomic PCR from IE35/BSC5 and w 1118 /BSC5 males, indicated that the 5Ј-end of the X chromosome fragment of the BSC5 duplication appears to be in the 3Ј-flanking region between Tis11 and Smr, because PCR products were not detected 1.8 kb downstream from the 3Ј-end of Tis11 mRNA but were detected in the last exon of Smr. Therefore, IE35/BSC5 males carry all of the genes lost in IE35 except for Tis11. These males will be referred to as Tis11 "null" flies. Tis11 null males were viable and fertile and appeared normal, although they exhibited a modest (ϳ24-h) delay in eclosion (data not shown). Given the extents of IE35 and BSC5, these males also carry two copies of ϳ13 genes distal to CG32650 6 (also see below).
We also used G1183 flies as a completely separate genotype that might lead to Tis11 depletion, presumably by the EP element interfering with transcription from the first exon. As described under "Materials and Methods," this mutation was initially lethal when it arrived in the laboratory but spontaneously reverted to a viable phenotype after several months of propagation (G1183v). We also generated a viable stock 6 K. Cook and E. Spana, personal communication. (G1183rec) by allowing meiotic recombination with w 1118 , as discussed under "Materials and Methods." The adult males of both G1183v and G1183rec were viable and fertile and expressed approximately one-eighth of the normal level of Tis11 mRNA (Fig. 5C); these alleles can therefore be viewed as stable hypomorphic alleles of Tis11. As shown in Fig. 5C, nei-ther Smr nor Ck1␣ expression was affected in either the spontaneous (G1183v) or recombinant (G1183rec) flies. We then back-crossed the G1183rec flies into the y w 67c23 background for 10 generations for the mRNASeq data experiments performed for this paper. As before, these flies were viable and fertile, although they also exhibited eclosion delays of ϳ24 h. showing the ratios of null/control and E showing the ratios of G1183/control (the centromere is to the right in both panels). In D, the location of genes deleted in IE35 and the genes duplicated in BSC5 are indicated at the bottom of the graph. Note the absence of Tis11 transcript expression in D as well as the depressed expression of several genes centromeric to Tis11 on the X chromosome. In E, the Tis11 transcripts were decreased by 8.6-fold, but most neighboring genes were relatively unaffected in the G1183 flies. DECEMBER 19, 2014 • VOLUME 289 • NUMBER 51

Drosophila Tis11
Deep Sequencing-Our working hypothesis was that depletion of Tis11 should lead to the abnormal accumulation of one or more of its target mRNAs. To search for potential target transcripts for Tis11 in adult flies, we first performed deep sequencing of transcripts (mRNASeq) on poly(A) ϩ -selected RNAs from null males. Values were compared with those obtained from w 1118 /BSC5 controls. We also performed a similar experiment with the G1183 males and their respective controls and compared these results with the results of the experiment with the null flies. The complete data set from these comparisons has been deposited in the Gene Expression Omnibus (accession number GSE58937).
Because the combined deletion/duplication approach for knocking out Tis11 was likely to lead to alterations in the expression of neighboring genes through the phenomenon of position-effect variegation (52), we evaluated levels of expression for the genes flanking Tis11 on the X chromosome. For this purpose, we used the read counts from the shared exons from all of the transcript variants for a given gene (i.e. a gene level comparison). We plotted these as the ratios of null/control average expression levels (Fig. 5D). We performed a similar analysis with the G1183 versus control experiment (Fig. 5E). In the Tis11 null flies, expression levels of the transcripts expressed by genes on the X chromosome that were distal (i.e. telomeric from) Tis11 were remarkably normal (Fig. 5D). As expected, Tis11 mRNA was not detected in the null flies. However, as predicted from the position-effect variegation concept (52), several genes close to Tis11 exhibited decreased expression in the null flies compared with controls; further away from the Tis11 locus, expression levels in the null flies returned to normal (Fig. 5D). We performed a similar analysis on the ratios of the average read counts for the G1183 flies compared with control (Fig. 5E). Other than Tis11 mRNA, which was downregulated by ϳ8.6-fold in these flies, the only significant change noted was a modest (ϳ40%) decrease in the expression of the immediately telomeric gene, Ck1␣, in the G1183 flies.
We next identified transcripts at the individual transcript variant level that were increased by 1.5-fold or more in both the Tis11 null flies and the G1183 flies, with p Ͻ 0.05 using unpaired Student's t tests. We identified 379 transcripts that were elevated in the null flies compared with controls using these criteria. In the G1183 experiment, there were 1430 upregulated transcripts, using the same criteria. When we compared the two data sets, we identified 69 transcripts, representing 56 genes, that were elevated in both experiments by at least 1.5-fold with p Ͻ 0.05. Table 1 shows the transcripts that were increased according to these criteria, along with their -fold increases in the Tis11-deficient flies compared with controls and the p and q values. We also searched the relevant 3Ј-UTRs from FlyBase for the presence of the core 5-mers for TTP family protein binding sites (AUUUA) as well as the longer 7-mer core binding site (UAUUUAU). Of the 69 transcripts, 34 (49%) contained at least one binding site 7-mer, whereas many more contained at least one core 5-mer. This compares with a frequency of 7-mers in the 3Ј-UTRs of Drosophila expressed transcripts of ϳ27% (data not shown); the difference between the two frequencies was significant by the hypergeometric test at 1.8EϪ05. Finally, we searched FlyBase for predicted cellular functions for each of the elevated genes (Table 1). A wide variety of potential functional classes of proteins was represented, but the most common cellular function listed as the primary functional term was proteolysis, listed as a primary defining term for 10 of the 69 jointly elevated transcripts.
A subset of the transcripts listed in Table 1 as well as some of the Tis11 isoforms were analyzed by NanoString nCounter analysis (46), using many of the original RNA samples used in the mRNASeq analyses. Only three of four of the null samples were available for this analysis. Of 42 up-regulated transcripts from Table 1 that were analyzed by this means, one was removed because the assay yielded results that were below the upper limit of the negative controls. Of the 41 remaining transcripts, eight were not significantly increased in either comparison, 12 were significantly up-regulated (p Ͻ 0.05 using unpaired, two-tailed Student's t tests) in one comparison, and 21 were significantly up-regulated in both types of mutant flies. The results from these confirmation assays are indicated by asterisks in Table 1, and a subset is shown graphically in Fig. 6 (see below).
As determined by mRNASeq analysis, there was no detectable expression of Tis11-RC, the longest transcript variant for Tis11, in the null flies, and this transcript was decreased 8.6-fold in the G1183 flies (Fig. 6A, left). Very similar changes in transcript levels were observed in the NanoString confirmation assays (Fig. 6A, right). In Fig. 6B, each panel represents the average depth of read coverage from the four biological samples used for each set, covering the 5Ј-end of the Tis11-RC mRNA. The arrow points to the site of insertion of the G1183 EP element. There were no detectable transcript reads 5Ј of the site of EP element insertion (Fig. 6B). Note that the vertical axis in the fourth panel (G1183) is 10-fold expanded compared with the other three. Shown in Fig. 6, C and E, are the means Ϯ S.D. of the expression results from two ect transcript variants, ect-RA and ect-RD, which have different 3Ј-UTRs. In both cases, these results were confirmed by the NanoString assays (right panels in Fig. 6, C and E. The alignments in Fig. 6, D and F, show sequences from the ect-RA and ect-RD transcripts compared with orthologous sequences from other Drosophila species contained in the "classical" tree, with the optimum binding heptamer UAUUUAU highlighted in red. Identical residues at each position are indicated by asterisks. In Fig. 6G are shown the transcript expression results for the mRNA encoding Rab18, a small GTPase, whose 3Ј-UTR contains only a single heptamer binding site that is less well conserved in some of the evolutionarily more distant Drosophila species (Fig. 6H). The increases in both mutants were confirmed by NanoString assays (of Fig. 6G, right). In Fig. 6, A, C, E, and G, all differences between the respective pairs of means were significant (p Ͻ 0.01).
We performed a similar analysis for the down-regulated transcripts. In the experiment comparing the null flies with controls, 623 transcripts were down-regulated more than 1.5fold with p Ͻ 0.05. In the experiment comparing G1183 flies with controls, 568 transcripts met the same criteria. When the data from the two experiments were combined, there were only 26 transcripts, representing 19 genes, in common between the two lists of mRNAs that were down-regulated by 1.5-fold or The data shown here are from the two mRNASeq experiments described under "Materials and Methods." For the column labeled "KO/Control," the results are expressed as -fold changes of transcript levels in the IE35/BSC5 "null" males divided by transcript levels in the w 1118 /BSC5 males used as controls. For the column labeled "G1183/Control," the results are expressed as -fold changes in transcript levels in the G1183 males over the y w 67c23 males used as controls. In each case, the Max-P and t test p values of the comparisons of the means from experimental versus control are shown. The column labeled "5-mers" shows the number of core 5-mers from the optimum TTP family member binding site, AUUUA, in the FlyBase annotated 3Ј-UTRs for each transcript; similarly, the column labeled "7-mers" shows the number of 7-mers, UAUUUAU, in the FlyBase annotated 3Ј-UTRs. The final column refers to the Gene Ontology molecular function or biological process from FlyBase; blanks indicate that no function or process was listed for that transcript. In the column "Gene name," a double asterisk indicates that both comparisons were statistically significant and in the same direction in the NanoString assay, and a single asterisk indicates that one of the two comparisons was significant; NC indicates "not confirmed" in either comparison; and a blank indicates that that transcript was not checked in the NanoString assay.

GeneAcc
Gene  Table 2). The largest changes were in the various isoforms of Tis11 mRNA. For Tis11-RC, the downregulation was 4370-fold for the null flies over control, whereas the down-regulation was 8.6-fold in the G1183 flies compared with the control flies (see Fig. 6A, left). These changes were confirmed by the NanoString assays. The Tis11-RC mRNA contains seven 7-mer potential Tis11 binding sites, scattered throughout its long 3Ј-UTR, suggesting that this transcript might itself be subject to autoregulation by its encoded protein.
With this exception, none of the other down-regulated transcripts shown in Table 2 contained 7-mers of the UAUUUAU type. of the mRNASeq expression results from two ect transcript variants, ect-RA and ect-RD, which have different 3Ј-UTRs; data from the NanoString analysis of a subset of these samples are shown in the right panels of C and E. The potential TTP family member binding sites in these transcripts are shown in D and F, respectively, and only show sequences that are conserved among the Drosophila species contained in the "classical" tree described in FlyBase. The optimum TTP binding heptamers, UAUUUAU, are highlighted in red. Identical residues at each position are indicated by asterisks. In the left panel of G are shown the mRNASeq expression data for Rab18 mRNA, encoding a small GTPase, whose 3Ј-UTR contains only a single heptamer binding site, with the corresponding NanoString data in the right panel of G. The potential TTP family member binding site in this transcript was less well conserved than seen with the other potential targets in this figure in some of the evolutionarily more distant Drosophila species (H). In A, C, E, and G, the mutant means (i.e. KO and G1183) from both the mRNASeq and NanoString analyses were statistically different from their respective WT means (p Ͻ 0.01 using Student's t test) in all cases.

DISCUSSION
We have begun to explore the functions of the Drosophila Tis11 protein in intact flies in order to unravel its role in the development and physiology of Drosophila and other insects. A long term goal is to use the advantages of the Drosophila system, including the fact that only a single TTP family member is expressed in this species, to help us understand the mechanisms of action of the entire TTP protein family in molecular and biochemical terms. We found that most of the biochemical characteristics of Tis11 were remarkably similar to those of its mammalian relatives. For example, when the protein was expressed in 293 cells or expressed in and purified from E. coli, it bound to the optimum target mRNA sequence as the mammalian proteins (i.e. UUAUUUAUU). This binding was dependent upon the presence of key cysteines and histidines within the TZF domain of the Tis11 protein and occurred with high affinity, with an apparent K d of 2.3 nM, remarkably similar to that seen with a synthetic TZF domain from the human protein (35). Whether this is the optimum binding sequence for the fly protein remains to be determined. The Tis11 protein also caused the deadenylation and destabilization of target transcripts co-expressed in 293 cells in a manner similar to its mammalian counterparts, an effect that was at least partially abrogated by removal of a C-terminal, highly conserved sequence that, in human TTP, forms a binding site for the NOT1 protein and presumably its associated multiprotein complexes (19,33), which include multiple deadenylases. Importantly, the same C-terminal sequence does not appear to contain the Drosophila nuclear export sequence (48), in contrast to what has been seen with some TTP family members in mammals (7,47), thus allowing these two functional domains of the protein to be experimentally separable. These findings suggest that the func-tions of Tis11 in Drosophila are very similar to the functions of vertebrate TTP family members (i.e. to bind to AU rich elements in the 3Ј-UTRs of transcripts and to promote their destabilization, apparently by initially stimulating deadenylation).
A remarkable finding was that Tis11 could promote the decay of target RNAs in transfection studies in human cells, at amounts of transfected DNA similar to those used with an effective amount of the human TTP expression plasmid. Based on the alignment between the Tis11 TZF domains with those of the mammalian proteins and the effect of amino acid mutations of key residues within the Tis11 TZF domain to completely abrogate RNA binding, we can assume that the Tis11 TZF domain was involved in binding to mRNA targets in the human cells. Surprisingly, not only was the fly protein expressed in human cells largely intact, but it apparently managed to utilize the human cellular machinery to effect mRNA processing. This suggests that the fly protein can recruit the relevant human proteins to assemble one or more deadenylation complexes, with an example being the involvement of the highly conserved C-terminal putative NOT1 binding region in Tis11-promoted mRNA decay. It is striking that the TZF domain and the putative NOT1 binding domain were the only regions of the protein that were highly conserved among insects and also were highly conserved with the human TTP family members. It seems possible that these two sequence regions, separated by non-conserved sequences, might be sufficient to promote TTPlike mRNA deadenylation and turnover.
To begin to elucidate the normal role of this protein in Drosophila physiology and development, we examined the effects of its loss of function on mRNA levels in adult male flies. The G1183 allele, in which an EP transposon was inserted into the first exon and interrupted the 5Ј-UTR, caused a decrease in Tis11 mRNA expression by more than 8-fold. G1183 homozygotes were viable and fertile. A

TABLE 2 Transcripts down-regulated in adult Tis11 null and hypomorphic males
The data shown here are from the two mRNASeq experiments described under "Materials and Methods." Column labels and other information are the same as in the legend to Table 1. Sdic1-RA, CG34166-RA, and CG40160-RA had two 5-mers, dob-RA had one 5-mer, and CG9317-RA had three 5-mers in their 3Ј-UTRs. Aside from Tis11-RC, none of these down-regulated transcripts contained UAUUUAU 7-mers in their 3Ј-UTRs. According to FlyBase, the longest Tis11 mRNA form, Tis11-RC, had a very long 3Ј-UTR (3317 bases) that contained 21 5-mers as well as six 7-mers, but the Tis11-RA isoform did not contain either motif. completely separate genetic model resulted in complete Tis11 deficiency, and these flies were also viable and fertile. These two independently derived Tis11-deficient models allowed for the identification of higher probability direct Tis11 targets than could be identified in either model alone. Neither of these models is optimal. In the case of the null flies, in which a large deletion in the X chromosome is partially compensated for by a duplication, Tis11 mRNA expression was completely abolished. Expression of the immediately flanking genes was normal, but there was diminished expression of several more distant neighboring genes, presumably by the phenomenon of position effect variegation (52). The G1183 allele resulted in an 8 -9-fold decrease in Tis11 mRNA levels but did not seem to affect the expression of the immediate flanking genes in adult males. Nonetheless, the significantly affected potential target transcripts that were seen in both fly stocks suggest that at least some of them are likely to be due to the Tis11 deficiency held in common by the two stocks, either directly or as downstream effects.

FlyBase ID
In each stock, large numbers of transcripts accumulated abnormally in the Tis11-deficient adults, suggesting their possible identity as Tis11 target transcripts. However, only 69 transcripts were elevated in both stocks at a level of 1.5-fold or more, and p Ͻ 0.05. Many of these were confirmed as increased in one or both stocks using a separate RNA quantitation assay. Half of the up-regulated transcripts contained potential TTP family member binding sites in their 3Ј-UTRs, and many of those were conserved among other Drosophila species. The likelihood is that some of these up-regulated transcripts are direct Tis11 targets, whereas others were elevated as a secondary consequence of Tis11 deficiency; future work will be necessary to distinguish between these two possibilities. Future experiments will also be necessary to reconcile the large number of transcript changes that occurred in each model that were not shared by the other model.
Recent developmental expression analyses have demonstrated that high levels of Tis11 transcripts were observed throughout embryogenesis, particularly very early and quite late, and in early adulthood. This information suggests that comparisons between the WT and Tis11-deficient strains during embryogenesis might be particularly informative. These analyses also found that in larvae, the highest expression level was seen in the trachea, whereas in adults, the highest level was seen in the virgin spermatheca. Unfortunately, despite the fact that our antiserum directed at the intact protein reacted strongly in Western blots using the recombinant protein at high antibody dilutions (e.g. 1:10,000), we have not succeeded in using this antibody to identify the endogenous Tis11 protein, either in Western blots of one-or two-dimensional gels or in tissue immunostaining. Further work will be necessary for the complete characterization of endogenous Tis11 protein expression patterns.
An interesting example of a potential direct target transcript is encoded by ect. The predicted protein from this gene was initially described as a mouse interferon ␣-like protein based on the similarity of the AU-rich region in the 3Ј-UTRs of the two sequences (53). However, the encoded proteins do not resemble interferon and have no known function. ect is known to be expressed during mid-embryogenesis, especially in the ectodermal tissues of the embryo (54); this pattern of expression was confirmed by recent analyses. The same report described negligible expression during adulthood in both males and females. In our studies, ect transcripts were present at low levels in the control flies but were strongly up-regulated in both types of Tis11-deficient flies. The various transcripts of this gene contain several ideal potential binding sites for Tis11, and many of these are conserved in ect orthologues from other Drosophila species. We propose that the Tis11 protein may play a role in destabilizing ect transcripts; it will very interesting to examine this possibility during developmental stages when both transcripts are relatively highly expressed (e.g. in embryos). Studies to examine the effects of Tis11 deficiency at various developmental stages are currently under way.
In previous work, Dorner et al. (26) used a genome-wide RNAi approach in cultured Drosophila cells to determine that Tis11 was a major component in the RNAi pathway in this system. Our data do not shed light on this possibility. More recently, three papers described a role for the Tis11 protein in the regulation of the ARE-containing mRNA encoding the CecA1 antimicrobial peptide after stimulation of the innate immune system (25,27,29). In our mRNASeq analysis of mRNAs from adults, the CecA1-RA mRNA was up-regulated 2.8-fold in the G1183 flies (p ϭ 0.03) compared with control, but it was not significantly elevated in the null flies compared with control. In the NanoString confirmation assay, there was a significant increase in the null flies but not in the G1183 flies. The CecA1-RA mRNA was fairly highly expressed in the adult males flies used here, with average RPKM values of 39.9 in the WT flies. For comparison, Tis11-RC had an average value of 17.6 in the same samples, and the 5th percentile value for this data set was 1.46. It will be of interest to examine the effects of Tis11 deficiency on the levels and turnover of this transcript after innate immune system stimulation in adult flies. A role for Tis11 in the decay of the transcript encoding eya (eyes-absent) was also described very recently in S2 cell experiments (30); however, we did not see significant increases in eya transcripts in either deep sequencing analysis of the Tis11-deficient adults. In contrast to CecA1-RA, the eya-RB transcript (the highest expressing transcript variant) was poorly expressed in the adults, with an average RPKM in the WT flies of 2.7.
Recently, Spasic et al. (28) examined the Drosophila transcriptome for AU-rich elements and used a dsRNA approach in Drosophila SL2 cells to search for transcripts up-regulated in response to Tis11 knockdown. They identified 53 transcripts that were up-regulated more than 1.41-fold in the Tis11-deficient cells. Remarkably, there was only overlap of a single transcript between these 53 and the 69 transcripts identified in the current experiments as up-regulated in both types of Tis11deficient adult flies. The common transcript was the vir-1 mRNA, which is induced in response to certain viral infections (55). It was significantly up-regulated by 1.63-fold in the null/ control comparison and by 1.85-fold in the G1183/control comparison. Its 3Ј-UTR contains two 7-mer potential binding sites that are conserved in several other Drosophila species. The difference in results when comparing our mRNASeq data with the microarray data of Spasic et al. (28) raises the possibility that the SL2 cells, derived from late stage Drosophila embryos, might be a more appropriate model for embryonic Tis11 activity than for the activity in the adult fly, a possibility that can be addressed in future experiments. However, we should emphasize that many other transcripts were up-regulated in one or the other of our comparisons but not both, and it seems likely that some of these overlap with the set of up-regulated transcripts identified by Spasic et al. (28).
Many mammalian ARE-binding proteins appear to have orthologues in Drosophila, but their potential interactions with Tis11 are not known. For example, the four AUF1 isoforms in humans (also known as heterogeneous nuclear ribonucleoprotein D0 proteins, or HNRPD proteins) appear to be orthologous to several isoforms of the Drosophila Squid proteins. We recently showed that two of the four human AUF1 isoforms can interact directly with the TZF domains of TTP family proteins, and this binding appears to potentiate the ability of TTP to bind to RNA (56). The TTP binding site within the AUF1 proteins is the so-called GY region near the C terminus; interestingly, several of the Drosophila isoforms have a similar GY-rich region in the same location, which is conserved in other insects. This raises the possibility that this particular interaction could be conserved between mammals and flies, a prospect that can be tested experimentally. Other ARE-binding proteins in flies include several isoforms of the ELAV (embryonic lethal abnormal vision) proteins, whose orthologues are the ELAV-like proteins 1-4 in humans (also known as HuA, HuB, HuR, etc. proteins in mammals). Both ELAV and HuD, when expressed ectopically in Drosophila, appeared to prevent the decay of AUrich element-containing transcripts that were also expressed ectopically (57). Chen et al. (58) showed that ELAV can bind to the 3Ј-UTR of diptericin mRNA; this gene is rapidly and transiently expressed following immune stimulation. ELAV was later shown to be involved in splicing of AU-rich containing introns (59). HuD was found to bind to AREs within the LIM only domain protein 4 (LMO4) mRNA in response to ATP, resulting in increased stability of the LMO4 mRNA (60). To our knowledge, neither the potential interactions of these AREbinding proteins with Tis11 nor their possible genetic interactions have been explored in Drosophila.
In conclusion, we have found that the biochemical properties of the Drosophila Tis11 protein are remarkably similar to those of its mammalian relatives, despite sequence conservation of only the RNA binding domains and the putative C-terminal NOT1 binding domains. In adult flies, two genetic models of Tis11 deficiency yielded many abnormally accumulating transcripts, many of which contained highly conserved potential Tis11 binding sites and which may well represent physiological targets for Tis11. Future studies will explore the effects of Tis11 deficiency on gene expression during various developmental stages in flies as well as the effects of infectious agents and other environmental stimuli on the possible Tis11 modulation of target transcript responses.