Interactions of CCCH Zinc Finger Proteins with mRNA

Macrophages derived from tristetraprolin (TTP)-deficient mice exhibited increased tumor necrosis factor α (TNFα) release as a consequence of increased stability of TNFα mRNA. TTP was then shown to destabilize TNFα mRNA after binding directly to the AU-rich region (ARE) of the 3′-untranslated region of the TNFα mRNA. In mammals and in Xenopus, TTP is the prototype of a small family of three known zinc finger proteins containing two CCCH zinc fingers spaced 18 amino acids apart; a fourth more distantly related family member has been identified inXenopus and fish. We show here that representatives of all four family members were able to bind to the TNFα ARE in a cell-free system and, in most cases, promote the breakdown of TNFα mRNA in intact cells. Because the primary sequences of these CCCH proteins are most closely related in their tandem zinc finger domains, we tested whether various fragments of TTP that contained both zinc fingers resembled the intact protein in these assays. We found that amino- and carboxyl-terminal truncated forms of TTP, as well as a 77 amino acid fragment that contained both zinc fingers, could bind to the TNFα ARE in cell-free cross-linking and gel shift assays. In addition, these truncated forms of TTP could also stimulate the apparent deadenylation and/or breakdown of TNFα mRNA in intact cells. Alignments of the tandem zinc finger domains from all four groups of homologous proteins have identified invariant residues as well as group-specific signature amino acids that presumably contribute to ARE binding and protein-specific activities, respectively.

Macrophages derived from tristetraprolin (TTP)-deficient mice exhibited increased tumor necrosis factor ␣ (TNF␣) release as a consequence of increased stability of TNF␣ mRNA. TTP was then shown to destabilize TNF␣ mRNA after binding directly to the AU-rich region (ARE) of the 3-untranslated region of the TNF␣ mRNA. In mammals and in Xenopus, TTP is the prototype of a small family of three known zinc finger proteins containing two CCCH zinc fingers spaced 18 amino acids apart; a fourth more distantly related family member has been identified in Xenopus and fish. We show here that representatives of all four family members were able to bind to the TNF␣ ARE in a cell-free system and, in most cases, promote the breakdown of TNF␣ mRNA in intact cells. Because the primary sequences of these CCCH proteins are most closely related in their tandem zinc finger domains, we tested whether various fragments of TTP that contained both zinc fingers resembled the intact protein in these assays. We found that amino-and carboxyl-terminal truncated forms of TTP, as well as a 77 amino acid fragment that contained both zinc fingers, could bind to the TNF␣ ARE in cell-free cross-linking and gel shift assays. In addition, these truncated forms of TTP could also stimulate the apparent deadenylation and/or breakdown of TNF␣ mRNA in intact cells. Alignments of the tandem zinc finger domains from all four groups of homologous proteins have identified invariant residues as well as group-specific signature amino acids that presumably contribute to ARE binding and protein-specific activities, respectively.
Zinc finger domains within proteins can mediate interactions with DNA, RNA, other proteins, and small molecules such as diacylglycerols. One relatively uncommon class of zinc finger proteins contains fingers of the CCCH type, in which three cysteines and one histidine are thought to coordinate a single atom of zinc. Members of a subclass of the larger family of CCCH zinc finger proteins contain two tandem zinc fingers consisting of CX 8 CX 5 CX 3 H, where X refers to variable amino acids, spaced exactly 18 amino acids apart. The prototype of proteins of this CCCH double zinc finger subclass is tristetra-prolin (TTP), 1 also known as TIS11 and Nup475. It was first identified as the product of an immediate early response gene in fibroblasts and other cells stimulated with insulin, serum, or phorbol esters (1)(2)(3)(4). The same stimuli that increase transcription of this gene also stimulate its rapid serine phosphorylation (5) and rapid nuclear to cytosol translocation (6) in fibroblasts.
TTP-deficient mice appear normal at birth but rapidly develop a wasting syndrome accompanied by erosive arthritis, dermatitis, alopecia, autoantibodies, and myeloid hyperplasia. Essentially all of these abnormalities were prevented by the injection of monoclonal antibodies specific for mouse tumor necrosis factor (TNF␣) (7). Macrophages derived from these mice exhibited enhanced TNF␣ release, accompanied by increases in TNF␣ mRNA levels (8). This was due to an increase in stability of TNF␣ mRNA in macrophages derived from TTPdeficient mice (9). These findings implicated TTP as an intracellular regulator of TNF␣ mRNA stability and thus of TNF␣ biosynthesis. More recently, we have shown that TTP deficiency has a similar effect on the stability of another mRNA containing a so-called class II AU-rich element (ARE), that encoding granulocyte-macrophage colony-stimulating factor (GM-CSF) (10). In this case, there was a marked stabilization of GM-CSF mRNA in bone marrow stromal cells derived from TTP-deficient mice compared with control cells, indicating that TTP is also a physiological regulator of GM-CSF mRNA stability and thus of GM-CSF secretion (10).
TTP can bind directly to the ARE of TNF␣ mRNA (9,11). The integrity of both zinc fingers was required for this direct protein-RNA interaction, because a single C 3 R mutation within the CCCH motif from either finger abolished the ARE binding activity of TTP (11). The same mutations abrogated the ability of TTP to destabilize the TNF␣ mRNA in intact cells (11). These studies indicated that TTP could bind directly to the TNF␣ ARE and destabilize the TNF␣ mRNA in a zinc fingerdependent manner, apparently by initially stimulating its deadenylation (11).
Aside from TTP, two other members of this subclass have been identified to date in mammals: cMG1 (TIS11b, ERF1, and Berg-36) (12)(13)(14)(15) and TIS11d (ERF2) (15,16). All three of these proteins contain the two typical CCCH fingers, spaced 18 amino acids apart, with the sequence RYKTEL or a variant leading into each finger. Proteins with nearly identical double zinc fingers spaced 18 amino acids apart have been identified in Drosophila and yeast (17)(18)(19). In addition to the Xenopus homologues of the three mammalian proteins described above, which all contain two tandem zinc fingers, we recently cloned a fourth cDNA encoding a Xenopus protein (XC3H-4) that contained two CCCH zinc fingers that were 18 amino acids apart and were preceded by the R(K)YKTEL sequence, as well as an additional more carboxyl-terminal pair of CCCH zinc fingers that were more closely spaced and lacked the lead-in R(K)YK-TEL sequence (20). This mRNA was expressed only in the oocyte, eggs, and early embryo; its mammalian homologue has not been identified, to our knowledge.
Although TTP appears to be a normal physiological regulator of TNF␣ and GM-CSF mRNA levels, secondary to its ability to bind to and promote the destabilization of their mRNAs, the physiological functions of the other vertebrate CCCH double zinc finger proteins are unknown. In the present study, we assessed the ability of these family members to bind to the ARE of the TNF␣ mRNA, and their ability to influence the stability of this mRNA in intact cells. Our data indicate that representatives of all four known members of this subclass of CCCH proteins exhibit TNF␣ ARE-binding activity, as assessed by cell-free mRNA cross-linking and gel shift assays. Members of at least three of the families can also exhibit TNF␣ mRNA-destabilizing capabilities in intact cell transfection assays. Somewhat surprisingly, both cell-free and intact cell activities were exhibited by both amino-and carboxyl-terminal truncated forms of TTP and by a 77-amino acid internal domain of TTP that contains both zinc fingers. Alignments of this tandem zinc finger (TZF) domain from all vertebrate members of this subclass revealed many invariant residues within this domain, as well as group-specific signature amino acids that may be important for the specific cellular functions of these proteins.
Expression Constructs-Human TTP expression constructs H6E⅐HGH3Ј and CMV⅐hTTP⅐tag were made as described (11). CMV⅐mTTP⅐tag, which contained the entire protein coding region of mouse TTP, was made using the same methods. CMV⅐hTTP (97-173)⅐tag, CMV⅐hTTP (1-173)⅐tag, and CMV⅐hTTP (97-326)⅐tag, which all contained the double zinc finger domain (amino acids 104 -166 from Ref. 21) and part of the protein sequence of human TTP, were made as described (11). CMV⅐hTTP (97-173)⅐tag contained essentially only the zinc fingers and seven flanking amino acids on both ends. CMV⅐hTTP (1-173)⅐tag contained amino acids 1-173, and CMV⅐hTTP (97-326)⅐tag contained amino acids 97-326 (the last amino acid) of human TTP.
CMV⅐CMG1⅐tag was made by inserting a PCR fragment containing the entire protein coding region of rat cMG1 (Ref. 13; bp 108 -1190 of GenBank TM accession number X52590), into the vector CMV⅐BGH3Ј/ pBSϩ. The template cMG1 for the PCR reaction was generously provided by Dr. K. D. Brown (AFRC Institute of Physiology and Genetics Research, Babraham, Cambridge, UK). The epitope tag derived from the influenza virus hemagglutinin protein (22) was attached to the last amino acid of the cMG1 protein as described (23).
CMV⅐XC3H-1⅐tag and CMV⅐XC3H-3⅐tag were made by inserting PCR fragments containing the entire protein coding region of the corresponding Xenopus cDNA clones (20), as well as the epitope tag fused to the last amino acid of each protein, into the vector CMV⅐BGH3Ј/pBSϩ. CMV⅐U2AF 35 was made by inserting a PCR fragment containing the entire protein coding region of the splicing factor U2AF 35 (24) into the vector CMV⅐BGH3Ј/pBSϩ. Plasmid pRSET B, which contained the entire coding region of U2AF 35 , was provided by Drs. B. R. Graveley and T. Maniatis (Harvard University, Cambridge, MA), and was used as a template in the PCR reaction.
CMV⅐mTNF␣ containing a NarI-XbaI fragment spanning bp 127-1325 of a mouse TNF␣ cDNA sequence (GenBank TM accession number X02611) was made as described (11). The mTNF␣ cDNA clone, provided by Dr. B. Beutler (University of Texas Southwestern Medical Center, Dallas, TX), contained an incomplete 3Ј-untranslated region that ended at bp 1325 (GenBank TM accession number X02611), with 33 adenylate residues attached to the last T. This sequence is shown in Fig. 1

Transfection of HEK 293 Cells, Northern Analysis, and Cytosolic Extract Preparation
HEK 293 cells were maintained in minimal essential medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Transient transfection of 1.5 ϫ 10 6 cells with CMV⅐hTTP⅐tag or with expression constructs containing the protein coding regions of the other CCCH zinc finger proteins or with vector pBSϩ alone in calcium-phosphate precipitates was performed as described previously (23,25), except that the transfection mixture was allowed to stay on the cells for 16 -20 h, and the glycerol shock step was omitted. When cells were co-transfected with CMV⅐mTNF␣ and CCCH protein expression constructs, human growth hormone expression plasmid pXGH5 (Nichols Institute Diagnostics, San Juan Capistrano, CA) was also co-transfected to monitor transfection efficiency.
24 h after the removal of the transfection mixture, samples were taken from the cell culture medium, and the human growth hormone released was assayed according to the manufacturer's protocol. Total cellular RNA was then harvested from the HEK 293 cells using the RNeasy system (Qiagen, Valencia, CA). Northern blots were prepared as described (2). Blots were hybridized to random-primed, ␣-32 P-labeled cDNA probes coding for various CCCH zinc finger proteins, including mouse TTP (2), Xenopus XC3H-1 and XC3H-3 (20), rat cMG1 (13), or splicing factor U2AF 35 (24). Blots were also hybridized with a ϳ1-kb NarI-BglII fragment of a mTNF␣ cDNA (11) and a ϳ 0.3-kb fragment of mouse cyclophilin cDNA (bp 166 -480; GenBank TM accession number X52803).
Cytosolic extracts were prepared from HEK 293 cells 24 h after the removal of the transfection mixture. The cells were incubated on ice for 20 min in a buffer consisting of 10 mM HEPES, pH 7.6, 3 mM MgCl 2 , 40 mM KCl, 5% (v/v) glycerol, 0.5% (v/v) Nonidet-P40, 2 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 8 g/ml leupeptin (lysis buffer). Lysis of the cells and maintenance of intact nuclei were carefully monitored by microscopy. The nuclei and cell membrane debris were removed by centrifugation at 16,000 ϫ g at 4°C for 15 min. Glycerol was added to the supernatant (cytosolic extract) to 20% (v/v), and the resulting extract was stored at Ϫ70°C.
Correct sequences of all plasmid inserts were confirmed by dRhodamine Terminator Cycle Sequencing (Perkin-Elmer, Foster City, CA). To label RNA transcripts with [␣-32 P]UTP (800 Ci/mmol), the above plasmids linearized with XbaI were used as templates, and the Promega Riboprobe in vitro Transcription Systems protocol was employed. The resulting products were precipitated with ammonium acetate and ethanol.
Cross-linking of Proteins to RNA-Cytosolic extracts prepared from HEK 293 cells transfected with CMV⅐hTTP⅐tag or other zinc finger protein expression constructs or vector (5 or 20 g of protein) were incubated with 1.5 ϫ 10 6 cpm of RNA probe in a 96-well plate at room temperature for 20 min in 20 l of lysis buffer (without protease inhibitors). Heparin and yeast tRNA were added to final concentrations of 2.5 g/l and 50 ng/l, respectively, for an additional 10 min. The 96-well plate was then placed on ice and irradiated at 254 nm UV light in a Stratalinker (Stratagene, La Jolla, CA) for 30 min at a distance of 5 cm from the light source. RNA not associated with protein was digested with 100 units of RNase T1 (Life Technologies, Inc) for 20 min at room temperature and further digested with 25 g of RNase A (Amersham Pharmacia Biotech) at 37°C for 15 min. The remaining RNA-protein complexes were analyzed by SDS-PAGE (12 or 16% acrylamide gel) followed by autoradiography.
Western Blotting-Cell extracts (5-50 g of protein) were mixed with 1 ⁄5 volume of 5ϫ SDS sample buffer (26), boiled for 5 min, and then loaded onto 12 or 16% SDS-PAGE gels. Western blotting was performed by standard techniques. Membranes were incubated in Tris-buffered saline/0.3% Tween 20 (TBS/T) with either polyclonal antiserum HA.11 (1:2, 500) or an antiserum to U2AF 35 (27). Incubation of the membranes with second antibody and development were performed as described (8). For some blots, 125 I-protein A (0.2 Ci/ml in TBS/T; Amersham Pharmacia Biotech) was used in place of second antibody.
RNA Electrophoretic Mobility Shift Assay-Cytosolic extracts prepared from HEK 293 cells transfected with either vector alone or expression constructs driven by the CMV promoter (5 or 20 g of protein) were incubated with 2 ϫ 10 5 cpm of RNA probe at room temperature for 20 min in 20 l of lysis buffer (without protease inhibitors). Heparin and yeast tRNA were added to final concentrations of 2.5 g/l and 50 ng/l, respectively, for an additional 10 min. RNA not associated with protein was digested with 100 units of RNase T1 (Life Technologies, Inc.) for 20 min at room temperature; the reaction mixture was then loaded onto a 6% nondenaturing acrylamide gel and subjected to electrophoresis at 250 V for 90 min, in 0.4ϫ Tris borate/EDTA buffer.

Expression of XC3H-4 Protein and Its Fragments in Xenopus Oocytes
Preparation of XC3H-4 RNA in Vitro-cDNAs encoding various regions of the XC3H-4 protein (20), full length (amino acids 1-276); amino acids 1-120 (containing the tandem CCCH zinc fingers of the TTPtype); and amino acids 121-276 (containing the second pair of CCCH zinc fingers), were inserted into the BglII cloning site of plasmid pSP64TEN (a gift from Dr. Douglas Melton, Harvard University, Cambridge, MA). The epitope tag derived from the influenza hemagglutinin protein (22) was attached to the last amino acid of each of the peptides as described (23). Correct sequence of the inserts was confirmed by dRhodamine Terminator Cycle Sequencing (Perkin-Elmer). The plasmids were linearized by XbaI digestion and were used as templates to synthesize RNA in vitro. The RNAs were prepared with the use of the mMESSAGEmMACHINE SP6 Kit (Ambion, Inc., Austin, TX) following the manufacturer's protocol.
Microinjection of Oocytes-Ovary was removed from adult Xenopus females (Xenopus I, Ann Arbor, MI), and stage VI oocytes were separated from the ovary and manually defolliculated. Oocytes were allowed to recover for 16 h at 18°C in buffer OR-2 (5 mM HEPES, pH 7.8, 82.5 mM NaCl, 2.5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 3.8 mM NaOH). Oocytes were injected with 30 -50 ng of mRNA and incubated for 24 h at 18°C. Oocyte cytosolic extracts were prepared as described above for 293 cell extracts.

RESULTS
The other two human members of the mammalian CCCH double zinc finger protein family, ERF1 (cMG1 and TIS11b) and ERF2 (TIS11d), share strikingly similar amino acid sequences in the TZF region with TTP ( Fig. 1, underlined). Although the carboxyl termini exhibit some sequence similarities, there are major differences between TTP and the other two proteins, whereas the amino acid sequences of ERF1 and ERF2 are more closely related to each other ( Fig. 1).
To determine whether these other two family members shared the ability of TTP to bind to the ARE region of the TNF␣ mRNA and destabilize it, ARE binding studies were performed using proteins expressed in HEK 293 cells, and co-transfection assays with the TNF␣ mRNA expression construct were performed in the same cell type. Besides the TTP expression constructs, the new expression constructs used were made from rat cMG1 (13), which is the rat homologue of mouse TIS11b, human ERF1, and Xenopus XC3H-2, and Xenopus XC3H-3 (20), which is the Xenopus homologue of mouse TIS11d and human ERF2.
Effects of TTP-related Proteins on TNF␣ mRNA-In the ex- pression studies described below, we used a previously described (11) TNF␣ expression construct, CMV⅐mTNF␣. This construct does not contain the full 3Ј-untranslated region of mTNF␣; instead, it ends at base 1325 (of GenBank TM accession number X02611) followed by 33 adenylate residues encoded by the cDNA. The expression of TNF␣ mRNA from this construct allowed the detection of both the adenylated and deadenylated forms of this mRNA in the presence of TTP (11). TTP has similar effects on the full-length mouse TNF␣ mRNA (data not shown), but for technical reasons involving size overlap with the 18 S ribosomal RNA, these are more difficult to quantitate.
The HEK 293 cells used in the transfection experiments do not express endogenous TNF␣ or TTP mRNA (11). In the following paragraph referring to Fig. 2, all quantitation of North-ern blot mRNA expression was determined by PhosphorImager and was corrected for transfection efficiency by human growth hormone secretion and for gel loading by quantitating endogenous cyclophilin mRNA levels. When these cells were co-transfected with CMV⅐mTNF␣ and a range of concentrations of the human TTP expression construct CMV⅐hTTP⅐tag, the mTNF␣ mRNA exhibited a characteristic expression pattern. At a low concentration of TTP DNA (0.005 g/dish of cells; Fig. 2A, lane 2), the total amount of mTNF␣ mRNA was reduced to ϳ40% of control in this experiment ( Fig. 2A, lane 1). When the amount of co-transfected TTP DNA was increased to 0.01 g of DNA, the total amount of hybridizable TNF␣ mRNA was further reduced to 23% of control ( Fig. 2A, lane 3). When 0.1 and 1 g of TTP were co-transfected, a smaller species of mTNF␣ mRNA was increased in intensity, whereas the upper band decreased markedly ( Fig. 2A, lanes 4 and 5). We have previously shown by RNase H experiments that the lower band is a deadenylated species of TNF␣ mRNA (11). At these higher concentrations of TTP DNA, the total accumulated hybridizable mTNF␣ mRNA was actually greater than that seen in the cells co-transfected with vector alone (Fig. 2A, lane 1). After correcting for transfection efficiency, the amounts of hybridizable mTNF␣ mRNA from cells co-transfected with 0.1 or 1 g of CMV⅐hTTP⅐tag were 254 and 481% of control, respectively. The apparent decrease in hybridizable mTNF␣ mRNA seen at the highest concentration of TTP DNA (1 g; Fig. 2A, lane 5) is likely to be due to the global inhibition of transcription seen at this level of TTP expression, as noted previously (11); this was reflected in a marked decrease in human growth hormone expression from these cells, accounting for the elevated "corrected" value for the sample shown in lane 5. These results are similar to those described previously (11), in which PhosphorImager values from four independent experiments were normalized for both transfection efficiency and gel loading and then averaged. In that study, an average decrease of TNF␣ mRNA to 17% of control was seen at 10 ng of CMV⅐hTTP⅐tag DNA; this value increased to 173 and 300% of control at 50 and 100 ng of DNA, respectively. The mechanism of the increased accumulation of the deadenylated species of TNF␣ mRNA seen at higher TTP expression plasmid concentrations is not known but is a consistent and highly reproducible finding (11).
In the same co-transfection experiment, we tested the ability of the two TTP-related proteins to destabilize TNF␣ mRNA and to promote the formation of the deadenylated species. When either CMV⅐CMG1⅐tag (a rat cMG1 expression plasmid, representing the cMG1/TIS11b/ERF1 proteins) or CMV⅐XC3H-3⅐tag (a Xenopus XC3H-3 expression construct, representing the TIS11d/ERF2 proteins) was co-transfected with CMV⅐mTNF␣ into 293 cells, each exhibited a similar pattern to TTP in influencing the accumulation of TNF␣ mRNA. With very low amounts of co-transfected CMV⅐CMG1⅐tag DNA, 0.005 and 0.01 g, the total hybridizable amounts of TNF␣ mRNA were decreased to 40 and 27% of control, respectively ( Fig. 2A, lanes  6 and 7). The accumulation of the smaller species of TNF␣ mRNA was obvious at 0.1 and 1 g of co-transfected CMV⅐CMG1⅐tag DNA; total hybridizable TNF␣ mRNA was 106 and 663% of control, respectively ( Fig. 2A, lanes 8 and 9). Expression of the Xenopus protein, XC3H-3, resulted in a similar pattern. Transfection of low concentrations of CMV⅐XC3H-3⅐tag DNA (0.005 and 0.01 g) caused a decrease in total TNF␣ mRNA to 57 and 49% of control, respectively ( Fig. 2A, lanes 10  and 11). At 0.1 g of CMV⅐XC3H-3⅐tag DNA co-transfection, the characteristic two sizes of TNF␣ mRNA were detected ( Fig. 2A,  lane 12; 78% of control), whereas at 1 g of CMV⅐XC3H-3⅐tag DNA, the deadenylated species of the mRNA accumulated to 461% of control ( Fig. 2A, lane 13). The bottom panel of Fig. 2A   FIG. 2. Effect of TTP-related proteins on TNF␣ mRNA stability. CMV⅐mTNF␣ (1 g/plate) was co-transfected into 293 cells with TTP or with the other CCCH zinc finger protein expression constructs or with vector alone. Total cellular RNA was harvested as described under "Experimental Procedures." Each gel lane was loaded with 10 g of total cellular RNA. Electrophoresis and Northern hybridization were performed as described under "Experimental Procedures." A, lane 1, vector alone (BSϩ, 5 g/plate); lanes 2-5, CMV⅐hTTP⅐tag (0.005, 0.01, 0.1, and 1 g of DNA/plate, respectively); lanes 6 -9, CMV⅐CMG1⅐tag (0.005, 0.01, 0.1, and 1 g/plate, respectively); lanes 10 -13, CMV⅐XC3H-3⅐tag (0.005, 0.01, 0.1, and 1 g/plate, respectively). B, lane 1, vector (BSϩ, 5 g/plate); lanes 2-5, CMV⅐mTTP⅐tag (0.005, 0.01, 0.1, and 1 g/plate, respectively); lanes 6 -9, CMV⅐XC3H-1⅐tag (0.005, 0.01, 0.1, and 1 g/plate, respectively); lanes 10 -13, CMV⅐ U2AF35 (0.005, 0.01, 0.1, and 1 g/plate, respectively). In both A and B, for all samples represented in lanes 2-13, vector DNA was also added to make the total amount of co-transfected DNA 5 g/plate. The Northern blots shown in the upper panels were probed with 32 P-labeled mTNF␣ and mouse cyclophilin cDNA probes. The two arrows labeled TNF␣ indicate the two species of TNF␣ mRNA discussed in the text. The arrow labeled Cyclo indicates endogenous cyclophilin mRNA. Identical RNA samples as in the upper panels were blotted and probed with 32 P-labeled mouse TTP, rat cMG1, and Xenopus XC3H-3 cDNA probes (indicated as CCCH in A), or mouse TTP, Xenopus XC3H-1 (indicated as TTP), and human U2AF35 cDNA probes (in B), and are shown in the lower panels. The position of the 18 S ribosomal RNA is indicated.
demonstrates the levels of expressed TTP, cMG1, and XC3H-3 mRNA in the same samples.
As a control, we used the human RNA splicing factor U2AF 35 (24). This protein contains two putative zinc fingers of the CCCH class, which, instead of being 18 amino acids apart, are widely separated by 116 amino acids and are also not preceded by the YKTEL lead-in sequence. The U2AF 35 protein is the smaller subunit of the essential splicing factor U2AF (24). Its heterodimeric complex with U2AF 65 is thought to be required for recognition of the 3Ј splice acceptor site in pre-mRNA splicing (for reviews see Refs. 28 and 29). U2AF 35 has also been shown to interact with other proteins involved in splicing, such as SC 35 , SF2/ASF, tra, and tra2, and was originally not thought to directly bind to RNA (27,30); however, this conclusion has been revised recently (31)(32)(33). Although endogenous U2AF 35 mRNA in 293 cells was readily detectable (Fig. 2B, lower panel), the transfection markedly increased the expression of this mRNA (Fig. 2B, bottom panel, lanes 12 and 13) and protein (data not shown). In this co-transfection experiment, the TNF␣ mRNA was not affected by the expression of increasing amounts of U2AF 35 (Fig. 2B, top panel, lanes 10 -13). This is in contrast to the effects of two other TTPs, mouse (2) and Xenopus (XC3H-1 (20)) (Fig. 2B, top panel, lanes 1-9), both of which behaved like their human counterpart in this assay.
Binding of CCCH Zinc Finger Proteins to the ARE of TNF␣ mRNA-Using the 70 base 32 P-labeled TNF␣ ARE probe (bases 1281-1350; GenBank TM accession number X02611), we performed UV cross-linking experiments with extracts from 293 cells that had been transfected with constructs expressing these CCCH zinc finger proteins driven by the CMV promoter. In addition to human TTP (Fig. 3A, lane 3), both rat cMG1 (lane 4) and Xenopus XC3H-3 (lane 5) were cross-linked by the TNF␣ ARE probe (Fig. 3A). Two major cross-linked proteins bands were formed with XC3H-3; one with an M r of ϳ48,000, presum-ably representing the intact protein, and one with an M r of ϳ32,000, presumably representing a proteolytic fragment. Human U2AF 35 protein, whose widely spaced zinc fingers lack the YKTEL lead-in sequence, did not form detectable complexes with the TNF␣ ARE probe (not shown). The identities of the endogenous 293 cell proteins that cross-linked to this probe, with M r values of ϳ85,000, ϳ70,000, ϳ46,000, and ϳ35,000 (see lane 6), are unknown; their possible relationship to other ARE binding proteins described in the literature has been discussed previously (11).
The ARE binding activity of the two TTP-related proteins was also tested in RNA mobility shift assays. As we have shown (11), the TNF␣ ARE probe formed three RNA-protein complexes (I, II, and III) with a control extract prepared from 293 cells transfected with vector alone (Fig. 3B, lanes 2 and 7). When an extract from cells transfected with CMV⅐hTTP⅐tag was used, new complexes were formed, whereas complexes I, II, and III seen in the extract from vector-transfected cells decreased or disappeared (Fig. 3B, lane 3). Formation of new RNA-protein complexes was likewise observed when extracts from 293 cells transfected with either CMV⅐CMG1⅐tag or CMV⅐XC3H-3⅐tag were used in the assay (Fig. 3B, lanes 4 and  5). Considerable radioactivity also remained in the gel wells when extracts containing these two proteins were used (Fig.  3B, lanes 4 and 5). Human U2AF 35 RNA splicing factor did not form detectable complexes with the TNF␣ ARE probe under these conditions (Fig. 3B, lane 6). Expression of these proteins from the expression constructs was readily detectable using antibodies to the epitope-tagged proteins (Fig. 3C, lanes 2, 4,  and 5).
Binding of TTP and Related Proteins to Mutant TNF␣ ARE Probes-We previously determined that the flanking A nucleotides of the AUUUA motifs in the TNF␣ ARE were essential for binding of TTP to the TNF␣ ARE (11). We next evaluated the ability of the TTP-related proteins to bind to mutant ARE probes.
We first compared a longer TNF␣ ARE probe (bases 1281 to 1350; GenBank TM accession number X02611) that contained seven AUUUA motifs to a probe (bases 1309 -1332) that contained only four of the AUUUA sequences (see bottom of Fig. 4). Using probe 1281-1350, cytosolic complexes of M r ϳ85,000, ϳ 70,000, ϳ46,000, and ϳ35,000 were seen when vector-transfected cell extracts were used; these were less apparent when the shorter probe 1309 -1332 was used (Fig. 4, compare lane 2 with lane 7 or lane 6 with lane 11). Extracts prepared from 293 cells transfected with the hTTP, cMG1, and XC3H-3 expression constructs all formed complexes with both long and short probes (Fig. 4, lanes 3-5 and 8 -10).
We next studied the binding specificity of TTP and its related proteins to a mutant of the short probe 1309 -1332. None of the proteins was able to form detectable complexes with probe 1309 -1332 (A/G) (bottom of Fig. 4, mutation sites indicated by triangles), a mutated ARE probe in which the flanking A residues in the AUUUA motif were substituted with Gs (Fig. 4,  lanes 13-17).
Characteristics of a Fourth CCCH Protein-We previously identified a fourth prospective family member of the CCCH TZF protein family in Xenopus (accession number AAD24210 (20)). This protein, XC3H-4, contained two zinc fingers spaced 18 amino acids apart that contained all of the hallmarks of the TZF domains from the three proteins discussed above; in addition, it contained two additional, more carboxyl-terminal CCCH zinc fingers, spaced more closely together and containing more degenerate lead-in sequences (20). Data base searches revealed sequence similarity to the amino-terminal portions of two CCCH proteins from zebrafish (34) and carp (35). Subsequent correction of the fish DNA sequences (CAA71245.2 for carp CTH1 and CAB55775.1 for zebrafish CTH1) showed apparent homology with the Xenopus sequence over the entire lengths of the proteins (Fig. 5). To our knowledge, mammalian homologues of these proteins have not been cloned to date.
Attempts to express the Xenopus member of this group, XC3H-4, or various subdomains, in 293 cells failed to yield significant levels of mRNA or protein. 2 However, significant expression of protein was readily achieved by injecting mRNA into Xenopus oocytes; extracts from these oocytes were then used in ARE cross-linking studies.
Using the 32 P-labeled TNF␣ ARE probe (bases 1281-1350; GenBank TM accession number X02611), we performed UV cross-linking experiments with extracts from Xenopus oocytes that had been injected with in vitro transcribed RNAs encoding the full length, the first half (amino acids 1-120 of accession number AAD24210), or the second half (amino acids 121-276) of the XC3H-4 protein. Similar to human, mouse, or Xenopus (XC3H-1) TTP expressed in 293 cells (Fig. 6A, lanes 3, 4, and  10), extracts prepared from oocytes injected with Xenopus XC3H-4 RNA that encoded the full-length protein were crosslinked by the TNF␣ ARE probe (Fig. 6A, lane 6), whereas no probe-protein complex was detectable when extracts of buffer (mock) injected oocytes were used (Fig. 6A, lane 5). When an extract from oocytes injected with RNA encoding the first 120 amino acids of XC3H-4 protein was used in the UV crosslinking assay, a probe-protein complex with an apparent M r of 15,000 was observed (Fig. 6A, lane 7). This complex formation is presumably due to the TZF domain that is related to the one in TTP (see Fig. 8B below). When an extract from oocytes injected with RNA encoding the second half of the XC3H-4 protein (amino acids 121-276) was used in the UV cross-linking assay, no probe-protein complex could be detected (Fig. 6A, lane  8). The second half of the protein contains a pair of CCCH zinc fingers that each have the internal spacing of the TTP-type zinc fingers, but the two fingers are separated by only 7 amino acids, and they lack the R(K)YKTEL lead-in sequence. Human U2AF 35 protein also did not form detectable complexes with the TNF␣ ARE probe (Fig. 6A, lane 9). Expression of the XC3H-4 protein and its fragments from the RNA-injected oocytes was readily detectable using antibodies to the epitope-tagged proteins (Fig. 6B, lanes 4 -6).
Interaction of TTP Fragments with TNF␣ mRNA-As shown in Fig. 1, the human proteins ERF1 and ERF2 exhibit much greater similarity in amino acid sequence with TTP within the TZF domain than in other regions of the proteins, suggesting that the common TZF domain may be the key component of these proteins that binds to and regulates the stability of AREcontaining mRNAs. We have already shown that the integrity of both zinc fingers is necessary for TTP binding to the TNF␣ ARE in cell-free assays and for destabilizing TNF␣ mRNA in cell transfection experiments (11).
We asked next whether the TZF domain alone was sufficient 2 W. S. Lai and P. J. Blackshear, unpublished data.  in lanes 2, 3, 7, 8, 13, and 14 contained 5 g of protein; lanes 4 -6, 9 -11, and 15-17 contained 20 g of protein) with 1.5 ϫ 10 6 cpm of probe, UV cross-linking, and RNase digestion were performed as described under "Experimental Procedures." Lanes 1, 12, and 18 (PЈ, 1.5 ϫ 10 6 cpm/sample), probe alone after digestion with RNase. Lanes 2,6,7,11,13, and 17 (BSϩ), extracts from 293 cells transfected with 5 g of vector plasmid. Lanes 3-5, Extracts from 293 cells transfected with 1 g/plate of plasmid CMV⅐hTTP⅐tag, CMV⅐CMG1⅐tag, or CMV⅐XC3H-3⅐tag, respectively; vector DNA was added to make the total transfected DNA 5 g/plate. Extracts described in lanes 3-5 were also used for lanes 8 -10 and 14 -16. The RNA-protein complexes were resolved by SDS-PAGE (12% gel) followed by autoradiography. The exposure time for the gel using probes 1281-1350 and 1309 -1332 was 4 h at Ϫ70°C, and it was 8 h for the gel using 1309 -1332 (A/G). The positions of molecular mass standards are indicated to the left of the gel. The sequences of the probes used are shown at the bottom; the adenosine residues mutated to guanosine residues in probe 1309 -1332 (A/G) are indicated by the triangles.
for TTP to interact with the TNF␣ mRNA. Three expression constructs were prepared that all contained the TZF domain; this spans amino acids 104 -166 of human TTP ( (21) Gen-Bank TM accession number M63625). CMV⅐hTTP (1-173)⅐tag contained amino acids 1-173; CMV⅐hTTP (97-326)⅐tag contained amino acids 97-326 (the carboxyl terminus of the protein); and CMV⅐hTTP (97-173)⅐tag contained the double zinc finger domain flanked by seven amino acids at each end.
When cell extracts from 293 cells expressing these protein fragments were used in UV cross-linking experiments using TNF␣ ARE probes, TTP and all of its fragments could be cross-linked to a longer probe containing the full ARE (bases 1281-1350) as well as a shorter probe containing only four AUUUA motifs (bases 1309 -1332) (Fig. 7A, lanes 1-10). Neither the full-length TTP protein nor any of its fragments was able to form a detectable complex with the mutant probe 1309 -1332 (A/G), in which the flanking As of its four AUUUA motifs were replaced by Gs (Fig. 7A, lanes 11-25).
In RNA mobility shift assays using the TNF␣ ARE 1309 -1332 probe and 293 cell extracts prepared from cells trans- fected with these TTP expression constructs, each of these TTP fragments, like the full-length protein, was able to form a probe-protein complex (Fig. 7B, lanes 1-6). When an epitope tag antibody was included in the mobility shift assay, all of these TTP fragments formed super-shifted complexes (Fig. 7B,  lanes 8 -11), whereas there was no super-shifted complex formation with extracts from vector-transfected 293 cells (Fig. 7B,  lane 7). Although the amount of mRNA and protein expression from these TTP fragment expression constructs was somewhat decreased relative to full-length TTP, readily detectable amounts of these fragments were seen by probing a Western blot with an antibody to the epitope tag (Fig. 7C).
We then tested the ability of these truncated forms of TTP to cause deadenylation and/or degradation of the TNF␣ mRNA in intact cells. When 293 cells were co-transfected with CMV⅐mTNF␣ and either CMV⅐hTTP (1-173)⅐tag (Fig. 7D, lanes  1 and 5-7) or CMV⅐hTTP (97-326)⅐tag (lanes 1, 8 -10), the TNF␣ mRNA exhibited the shortening to the deadenylated form seen with full-length TTP (Fig. 7D, lanes 1-4). When construct CMV⅐hTTP (97-173)⅐tag was used, the characteristic two bands of TNF␣ mRNA were seen at both 1 and 5 g of transfected DNA (Fig. 7D, lanes 11-13). These data indicated that the 77-amino acid peptide containing the TZF domain alone was capable of promoting the decrease in size of the TNF␣ mRNA, which we have attributed to its deadenylation (11). The apparent differences in potency of these constructs to promote deadenylation of the TNF␣ mRNA appear to be due, at least in part, to differences in expression (Fig. 7, D, lower Fig. 4. B, RNA electrophoretic mobility shift assays using TNF␣ ARE probe 1309 -1332. Extracts (10 g of protein) from 293 cells transfected with full-length or truncated human TTP as described above were incubated with 2 ϫ 10 5 cpm of probe in the absence (Ϫ) or presence (HA.11) of a polyclonal anti-epitope tag antibody. RNA mobility shift assays and RNase T1 digestion were performed as described under "Experimental Procedures." Lane 1 (PЈ), probe alone (RNase T1 digested). Lanes 2 and 7 (BSϩ), extracts from 293 cells transfected with 5 g/plate of vector plasmid. Lanes 3 and 8, extracts from 293 cells transfected with 1 g/plate of plasmid CMV⅐hTTP⅐tag (full). Lanes 4 and 9, extracts from 293 cells transfected with 1 g/plate of plasmid CMV⅐hTTP(1-173)⅐tag. Lanes 5 and 10, extracts from 293 cells transfected with 1 g/plate of plasmid CMV⅐hTTP(97-326)⅐tag. Lanes 6 and 11, extracts from 293 cells transfected with 1 g/plate of plasmid CMV⅐hTTP(97-173)⅐tag. Vector DNA was added to make the total transfected DNA 5 g/plate. The super-shifted (SS) RNA-protein complexes are indicated. The RNA-protein complexes I, II, and III formed with endogenous 293 cell proteins and the migration position of free probe (FP) are indicated. C, Western analysis of full-length and truncated TTP protein expression. Cell extracts (5 g of protein) prepared from 293 cells transfected with 1 g of DNA from the constructs described above were separated on a 16% SDS-PAGE gel and immunoblotted using a polyclonal antibody to the epitope tag of the fusion proteins. Antibody was visualized with 125 I-protein A, and an autoradiograph of this blot is shown. Molecular mass standards are indicated to the left of the blot. D, effect of truncated TTP fragments on TNF␣ mRNA stability. CMV⅐mTNF␣ (1 g/plate) was co-transfected into 293 cells with either full-length or truncated human TTP expression constructs or vector alone. Total cellular RNA was harvested as described under "Experimental Procedures." Each lane was loaded with 10 g of total RNA. Electrophoresis and Northern hybridization were performed as described under "Experimental Procedures." Lane 1, vector (BSϩ, 5 g/plate); lanes 2-4, CMV⅐hTTP⅐tag (full) (0.1, 1, and 5 g/plate, respectively); lanes 5-7, CMV⅐hTTP(1-173)⅐tag (0.1, 1 and 5 g/plate, respectively); lanes 8 -10, CMV⅐hTTP(97-326)⅐tag (0.1, 1 and 5 g/plate, respectively); lanes 11-13, CMV⅐hTTP(97-173)⅐tag (0.1, 1 and 5 g/plate, respectively). Vector was added to each transfection mixture to make the total amount of co-transfected plasmids 5 g/plate. The Northern blot was probed with a 32 P-labeled mTNF␣ cDNA (upper panel). The two arrows to the right of upper panel indicate the two species of TNF␣ mRNA discussed in the text. Identical RNA samples as in the upper panel were blotted and probed with a full-length 32 P-labeled mouse TTP cDNA, as shown in the lower panel. The position of the 18 S ribosomal RNA is indicated. and C); this conclusion is supported by more extensive concentration-response experiments (not shown).
Alignment of TZF Domains of Known CCCH Proteins-To begin to identify critical sequence requirements for ARE binding, we have begun an analysis of the TZF domain of the vertebrate CCCH proteins. We aligned the TZF domains from the four vertebrate proteins discussed here, making the assumption (borne out in every case in which it has been tested directly) that the domains from homologues from other animal species would bind to the ARE probe similarly to the prototype protein. The TZF domains, with one exception, are from published sequences already in GenBank TM , as listed in the legend to Fig. 8 were aligned using the program Pileup from GCG. In Fig. 8A is illustrated the dendrogram produced by these alignments; despite the facts that these proteins are much more disparate outside of the TZF domains than within them, the alignment program still aligned them into homologous groups (Fig. 8A). The alignment itself is pictured in Fig. 8B. To simplify the discussion of the alignment, the amino acids within the TZF domains are numbered from 1 to 64. Although similar TZF domains have been identified in invertebrate animals and in plants, we excluded these sequences from the alignment because we have no direct evidence at present that they are capable of ARE binding or mRNA destabilization.
Examination of Fig. 8B reveals that 34 of the 64 amino acids in the TZF domains (53%) have been conserved among all four proteins from species as diverse as humans, Xenopus, and zebrafish. These include the RYKTEL lead-in sequence for the first zinc finger, the lead-in sequence KYKTEL in the second zinc finger, and several other amino acids in the inter finger 18-amino acid spacer, including a Gly residue at position 27, an acidic residue at 30, a Leu at 31, a His at 37, and a Pro at 38. Within both zinc fingers, the canonical CCCH residues were conserved. Within the first finger, a Glu residue was conserved at position 12, a Gly at 14, a Tyr at 18, a basic residue at 21, and Gln-Phe-Ala at 23-25. Within the second finger, a Gly residue was conserved at position 52, a Tyr at 56, an Arg at 59, a Phe at 62, and a branched chain amino acid at 63.
The TZF domains also appeared to contain protein-specific "signatures," which allowed the domains from a given protein (e.g. TTP) to be grouped appropriately with its homologues (Fig. 8A), despite the great evolutionary distance between the animal species examined. For example, TTPs from human, cow, rat, mouse, and Xenopus all contain Thr residues at position 9, Ser residues at position 11, Ala residues at position 20, Asn residues at position 35, His-Lys at 46 and 47, Tyr-Leu at 49 and 50, and Ser residues at position 58, differentiating the TTP homologues from all other proteins examined. Similarly, the cMG1 homologues all contain Asp residues at position 20 and Ile residues at position 28, distinguishing them from the others. The TIS11D proteins had as their only signature a Phe residue at position 28. The more distantly related XC3H-4 proteins also contained signature amino acids, including Ser-Arg at 8 and 9, Ala at 11, and Leu at 61.
These comparisons also identified some positions within the TZF domains that can tolerate significant amino acid diversity. For example, Pro, Thr, and Arg were present at position 9, Glu and Ser at position 11, Asn, Ser, and Thr at 13, Ala, Ser, Thr, Arg, and Phe at 15, Asp, Glu, Ala, and Asn at 20, Ile, Phe, Leu, Pro, and Lys at 28, His, Gly, Ile, and Ser at 29, etc. Less diversity is evident at other positions; for example, only hydrophobic residues were present at positions 10, 48, and 49, and only basic residues were present at position 21. DISCUSSION We recently showed that TTP, the prototype of a small family of vertebrate proteins with tandem CCCH zinc fingers, exhibited several measurable activities in both cell-free and intact cell experiments. In cell-free experiments, TTP bound and could be cross-linked to the ARE of the TNF␣ mRNA, as determined by both gel shift and UV light cross-linking experiments; this interaction could be prevented by mutating a single cysteine to arginine in either zinc finger (11). In cell transfection studies, TTP promoted the apparent deadenylation and breakdown of a 3Ј-truncated TNF␣ mRNA that contained most of the native ARE (11). That these effects of TTP are likely to be physiologically relevant has been demonstrated by experiments in TTP knockout mice, which are characterized by a chronic inflammatory state that appears to be largely due to excess circulating TNF␣ (7)(8)(9). This excess TNF␣ was derived, at least in part, from macrophages, because macrophages from the TTP knockout mice overproduced TNF␣. This in turn was due to an increased stability of the TNF␣ mRNA in the TTPdeficient cells (7)(8)(9).
More recently, we found that the GM-CSF mRNA, which contains a similar ARE in its 3Ј-untranslated region, was also greatly stabilized in bone marrow stromal cells derived from the TTP knockout mice (10). Importantly, these studies also identified an apparently deadenylated intermediate breakdown form of the GM-CSF mRNA in the cells from normal mice, whose formation was greatly inhibited in the cells derived from the TTP-deficient mice (10). Taken together with earlier data, these studies suggest that TTP plays an important physiological role in the destabilization of two important, AREcontaining mRNAs, those encoding TNF␣ and GM-CSF, and that this TTP effect appears to involve mRNA deadenylation.
It appears likely that the first step in the effect of TTP to destabilize these mRNAs is its binding to the ARE, because nonbinding zinc finger mutants do not stimulate TNF␣ mRNA breakdown in intact cells (11). The experiments described here begin to explore the structure-function requirements for this first step in this protein-RNA interaction. First, we showed that the two other known mammalian members of this class of tandem CCCH zinc finger proteins exhibited similar activities in our assays. Specifically, the rat cMG1 protein, whose homologues include mouse TIS11b, human ERF1, and Xenopus XC3H-2, and the Xenopus XC3H-3 protein, whose homologues include mouse TIS11d and human ERF2, stimulated the apparent deadenylation and destruction of the TNF␣ mRNA in intact cell transfection experiments. Second, these proteins, as well as a TZF domain from the fourth known vertebrate member of this protein class, Xenopus XC3H-4, could bind to the ARE of the TNF␣ mRNA in cell-free experiments, as demonstrated by gel shift and UV light cross-linking experiments. Finally, we found that a 77-amino acid TTP fragment containing the TZF domain of TTP was sufficient to mediate the TTP effect on TNF␣ mRNA stability in cell transfection experiments, as well as in cell-free gel shift and cross-linking experiments.
The three mammalian and one Xenopus/fish CCCH proteins studied here comprise the four known members of a subclass of vertebrate CCCH proteins in which the following features characterize the TZF domains: 1) Both fingers within the TZF domain in the four proteins are preceded by a conserved 6-amino acid lead-in sequence, R(K)YKTEL; 2) Both fingers in all proteins contain the following conserved residues and spacing, CXX(F/Y)X 3 GXCXYXX(K/R)CXFXH, where X represents variable amino acids; 3) Both fingers in all proteins are separated by exactly 18 amino acids, i.e. between the terminal H of the first finger and the first C of the second finger. These characteristics are identical in the protein homologues despite their species of origin, ranging from human to Xenopus laevis; 4) The three mammalian proteins and their homologues are basic, with overall pIs ranging from 8.75 to 9.91. In contrast, the Xenopus protein XC3H-4 has an overall pI of 5.9; however, the TZF domain itself is basic, with a pI of 9.1, similar to the pIs of the TZF domains from the other three proteins (20).
In contrast, the hU2AF 35 protein exhibited no activity in our assays. This smaller subunit of the essential mRNA splicing factor hU2AF was not thought to bind to RNA directly (24,30); however, three recent reports suggest that direct binding to RNA can occur (31)(32)(33). This protein contains two related CCCH zinc fingers that are present in the protein homologues from man to yeast but that are separated by 116 amino acids in the human protein. Thus, the mere presence of two zinc fingers of this type does not appear to confer TNF␣ mRNA destabilizing and ARE binding activity, in the absence of the other characteristics of the more closely spaced TZFs of TTP and its relatives. However, it will be of interest to determine whether the CCCH zinc fingers are involved in direct contacts with the pre-mRNA in this case.
The fact that TTP and its two related proteins all exhibited similar mRNA binding and destabilizing activities in our assays suggested that the domain they all hold in common, the TZF domain, is critical for these activities. That the TZF domain is necessary for these activities was demonstrated by our findings that mutation of a single cysteine in either zinc finger to an arginine (11) completely abrogated ARE binding and mRNA destabilizing activities of TTP. The present experiments demonstrate that a 77-amino acid domain of TTP that contains the TZF domain was also sufficient for both ARE binding and TNF␣ mRNA destabilizing activities. Therefore, at present, this 77-amino acid domain is the minimum known TTP sequence required for these activities. The corresponding minimum known ARE sequence required for this binding is 24 bases (bases 1309 -1332; GenBank TM accession number X02611).
These findings raise the possibility that these related proteins may have TTP-like activities in normal physiology, perhaps to regulate TNF␣ mRNA stability or perhaps to regulate the stability of other mRNAs containing class II or related AREs. The physiological importance of these findings will probably require the generation of knockout mice for each gene.
Many questions remain to be answered about these proposed interactions, particularly concerning the mechanisms of the effect of the CCCH proteins on mRNA stability as well as their specificity and physiological relevance. Among the most important is what confers the specificity of CCCH protein binding to Class II AREs, because all three CCCH proteins can be expressed in the same cell types, such as macrophages? There are major differences in the patterns of expression of the three proteins in the mouse, particularly in their biosynthetic responses to external stimuli and their tissue-and cell-specific expression. 4 These differences, as well as the markedly different primary sequences among the three proteins outside of the TZF domains, may help to explain why the endogenous mouse equivalents of cMG1 and XC3H-3 do not prevent the development of the TTP deficiency syndrome in mice and, in fact, were present in normal amounts in tissues of these animals (7). There are many previous examples in the literature in which disruption of the gene encoding a single member of a gene family is not compensated by the remaining members (36,37).
Another important question is that because the TZF domain alone seems to be adequate for at least the TTP activities measured here, what are the functions of the other domains in the three proteins? There are rapid, mitogen-induced changes in the biosynthesis, subcellular localization, and serine phosphorylation of TTP. What role do these modifications play in the behavior of TTP, both in cells expressing TNF␣ and in cells not expressing this cytokine? Do similar modifications of the other two proteins take place in certain cells and tissues with various stimuli? The availability of the cell-free and transfection assays described here, coupled with the knockout mouse approach to physiological function, should make it possible eventually to answer these questions.