INTRODUCTION

Tumor necrosis factor- (TNF-)¹ plays a pivotal role in inflammation and host defense (1). TNF- stimulates effector cell microbicidal and tumoricidal activity, enhancing survival of the infected host (2, 3) and causing regression of some tumors (4). On the other hand, persistent or inappropriately high TNF- expression has grave consequences, including multiorgan failure and death (5, 6). Thus, stringent control of TNF- regulation is critical for host survival and has led to the evolution of multiple levels of macrophage TNF- regulation (7).

The TNF- gene lies between the lymphotoxin- and lymphotoxin- genes in a tightly linked array, but these genes are regulated by distinct promoters (8, 9). TNF- transcription occurs in plastic-adherent but otherwise unstimulated macrophages in the absence of detectable translation (10, 11). The TNF- promoter shares several regulatory elements with other genes, including, but not limited to, four B-like elements, a GC box/Sp-1 binding site, a Y-box like decanucleotide, a cAMP-responsive element-binding protein binding site, and binding sites for AP-1 and AP-2 (12). Several macrophage activators induce a modest increase in TNF- transcription (13), which is blocked by agents that enhance intracellular cyclic AMP levels (14). In the ANA-1 macrophage cell line, TNF- transcript elongation is blocked in unstimulated cells. Bacterial endotoxin lipopolysaccharide (LPS) stimulation both increases TNF- transcription initiation and reverses the transcript elongation block (15). Once transcribed, TNF- mRNA is stabilized by the same stimuli that induce TNF- release in macrophages, including LPS (16). Enhancement of transcription initiation, release of the transcript elongation block, and transcript stabilization contribute to the accumulation of TNF- mRNA in stimulated macrophages. The TNF- mRNA that is present in unstimulated macrophages is translationally silent. Macrophage activation induces both accumulation and translation of TNF- mRNA (11, 13, 17). Cis acting elements in the TNF- 3-untranslated region are sufficient for both translational silencing in unstimulated macrophages and the LPS-induced activation of TNF- translation (17).

We (18) and others (19) have reported the translationally silent TNF- mRNA in unstimulated macrophages migrates faster than the translationally active TNF- transcripts in LPS-stimulated macrophages in Northern blot analyses. Possible causes of the observed differences in TNF- transcript size in unstimulated and stimulated macrophages include: 1) isoforms of the TNF- gene; 2) use of an alternative poly(A) signal site(s); 3) mRNA splice variants in unstimulated macrophages lacking one or two internal exons; or 4) heterogeneity in poly(A) tail length. Current data do not support the first two possible possibilities. Previous searches of genomic libraries did not reveal any genes with homology to TNF- (8) and only a single AAUAAA canonical nuclear poly(A) signal site is present in mouse and human TNF- sequences (20).

The objectives of this study are to determine the cause of the TNF- mRNA size heterogeneity in unstimulated and LPS-stimulated macrophages and to evaluate its relevance to the regulation of TNF- expression. We show that the TNF- mRNA in adherent, unstimulated monocytes and macrophages is approximately 200 nucleotides shorter than those that appear after LPS stimulation. The size difference is due to markedly shorter or absent 3 poly(A) tails on the TNF- mRNA in the unstimulated macrophages. We show that the 3 poly(A) tail shortening is not caused by either ongoing translation or dissociation of TNF- mRNA from polysomes and suggest that 3 poly(A) tail metabolism may play a critical role in TNF- regulation.

EXPERIMENTAL PROCEDURES

Phenol-extracted LPS prepared from Escherichia coli 0111:B4 was obtained from Difco. Unless otherwise specified, all other reagents were obtained from Sigma.

RAW 264.7 (RAW), J774A.1, and PU5-1R mouse macrophage cell lines were obtained from the American Type Culture Collection (Rockville, MD). The P388D₁ mouse macrophage cell line was the kind gift of Dr. R. Nordan (National Institutes of Health, Bethesda, MD). All cell lines were maintained in either RPMI 1640 (Mediatech; Fairfax, VA) supplemented with 50 units/ml penicillin, 50 µg/ml streptomycin, 2 mM L-glutamine, 1 mM sodium pyruvate, 10 mM Hepes buffer, pH 7.3 (Life Technologies, Inc.) (CRPMI), containing heat-inactivated 10% newborn calf serum (Hyclone; Logan, UT) or in Macrophage-SFM^TM (Life Technologies) serum-free medium without supplementation (SFM) at 37 °C in 5% CO₂ enriched air. Cells were routinely tested for Mycoplasma sp. infection using a commercial assay system (MycoTect; Life Technologies), and new macrophage cell line cultures were established monthly from frozen stocks. All media and reagents contained <0.1 ng/ml endotoxin as determined by a Limulus polyphemus amebocyte lysate assay (Associates of Cape Cod; Falmouth, MA). Whole blood was obtained from healthy volunteers following informed consent using a protocol approved by the University of Maryland at Baltimore Institutional Review Board. Peripheral blood mononuclear cells were isolated by Ficoll-hypaque density centrifugation (21), resuspended in CRPMI containing 10% autologous human serum, and cultured in Teflon beakers (Nalgene; Rochester, NY) for 7 days before use (22).

Total RNA was isolated by an acidic guanidinium thiocyanate-phenol-chloroform extraction as described (23). Cytoplasmic RNA was obtained by lysing cells in polysome buffer (PB) containing 10 mM Tris, pH 7.3, 150 mM NaCl, 10 mM MgCl₂, 0.65% Nonidet P-40 and collecting the postmitochondrial supernatant fraction after centrifugation at 10,000 × g for 15 min at 4 °C. The postmitochondrial supernatants were either applied directly to sucrose gradients for polysome separations or were extracted with SDS/phenol/chloroform/isoamyl alcohol followed by ethanol precipitation of the cytoplasmic RNA. Precipitated RNA was washed in ethanol, dissolved in water, quantitated by absorbance at 260 nm (A₂₆₀), and stored at 80 °C (24).

Ten micrograms of denatured RNA/lane as well as a RNA molecular size standard (Life Technologies) were separated by electrophoresis through either 1.5 or 3% agarose (SeaKem Gold, FMC Bio Products; Rockland, ME), 0.74 M formaldehyde (Fisher) gels. The separated RNA was transferred to nylon membrane (Duralon-UV, Stratagene; La Jolla, CA) by positive pressure (PosiBlot, Stratagene), UV-cross-linked (UV Stratalinker, 120,000 µJ, Stratagene), prehybridized in QuikHyb (Stratagene) for 15 min at 68 °C, and then hybridized for 1 h at 68 °C in QuikHyb containing 0.67 µg/ml denatured salmon testes DNA and ³²P-labeled cDNA probe. Following washes of increasing stringency, membranes were analyzed with a PhosphorImager (Molecular Dynamics; Sunnyvale, CA) and autoradiographed. These blots were stripped and reprobed with a second radiolabeled cDNA where indicated according to standard methods (24). Equal RNA loading was determined by UV visualization and photography of 18 and 28 S ribosomal RNA bands in ethidium bromide-stained gels and Northern blots as well as quantification of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) housekeeping gene transcripts on Northern blots.

A 0.8-kb cDNA probe for human GAPDH was obtained from M. Olman (University of California; San Diego, CA). A 2.0-kb cDNA for chicken -actin was obtained from D. Cleveland (Johns Hopkins University School of Medicine; Baltimore, MD). A 1.6-kb cDNA for mouse TNF- was provided by B. Sherry (Picower Institute; Manhasset, NY). A fragment of mouse TNF- cDNA comprising exon 4 was provided by S. Nedospasov (National Cancer Institute; Frederick, MD). For Northern blot analysis, cDNA inserts were purified and labeled with ³²P using a random primer kit (Life Technologies) according to the manufacturer's instructions.

Twenty micrograms of cytoplasmic RNA was denatured in a reaction mixture containing 0.6 µg oligo(dT)_12-18 (Life Technologies), 20 mM Tris, pH 7.5, 100 mM KCl, 10 mM MgCl₂, 0.1 mM dithiothreitol, and 5% sucrose at 70 °C for 5 min. The mixture was placed on ice for 15 min before the addition of 2.5 units RNase H (Life Technologies) and incubation at 37 °C for 30 min. The digest was extracted with phenol/chloroform/isoamyl alcohol followed by sodium acetate/ethanol precipitation of the RNA. Precipitated RNA was washed in ethanol, dissolved in water, quantitated, and stored at 80 °C.

Heat-denatured RNA was reverse transcribed in a reaction mixture containing 50 mM Tris, pH 8.3, 75 mM KCl, 3 mM MgCl₂, 0.5 µg of random hexamers (Life Technologies), 0.5 mM dNTP mix (Pharmacia Biotech Inc.), 0.01 mM dithiothreitol (Life Technologies), 40 units of RNasin (Promega; Madison, WI), 2 µg of acetylated bovine serum albumin, and 200 units of Moloney murine leukemia virus reverse transcriptase (Life Technologies) at 37 °C for 1 h. The reaction was terminated by heating at 95 °C for 5 min, chilled on ice, and stored at 20 °C. One microliter of the first strand product was amplified using 12.5 µM primer pairs and 0.625 units Taq DNA polymerase (AmpliTaq, AS; Perkin-Elmer) in a reaction mixture containing 2.5 mM MgCl₂, 10 mM Tris, pH 8.3, 50 mM KCl, and 0.5 mM dNTP mix (Pharmacia). Each cycle consisted of a 30-s denaturation at 95 °C, 1 min of annealing at 54 °C, and 3 min of extension at 72 °C. A total of 35 cycles were performed followed by a 10-min terminal extension (RoboCycler, Stratagene). Amplification products and a DNA molecular size standard (Life Technologies) were analyzed by electrophoresis in 1.5% agarose (Life Technologies) with UV visualization and photography. PCR primers were constructed: upstream sense primer, 5-TCAGCGAGGACAGCAAGGGA-3; downstream antisense primer, 5-CCAAGCGATCTTTATTTCTCTCA-3.

Twelve-milliliter 15-50% linear sucrose gradients in PB lacking Nonidet P-40 were prepared, and approximately 0.5 ml of postmitochondrial supernatant was layered on the gradient and centrifuged for 1 h, 45 min at 39,700 rpm (270,000 × g) in a SW41Ti rotor (Beckman Instruments, Inc.) at 4 °C. Fractions were collected, using a density gradient fractionator (Isco; Lincoln, NE; Brandel Inc.; Gaithersburg, MD), into equal volumes of isopropyl alcohol while monitoring A₂₅₄. RNA in the fractions was pelleted by centrifugation, extracted sequentially in acidic guanidinium thiocyanate (23) and phenol/chloroform/isoamyl alcohol, precipitated with isopropyl alcohol, washed in ethanol, dissolved in water, quantified by A₂₆₀, and analyzed by Northern blotting.

TNF- Protein Measurement

TNF- protein was measured by sandwich enzyme-linked immunosorbent assay using commercial reagents (Minikit, Endogen; Cambridge, MA) and expressed in pg/ml as calculated from a recombinant TNF- standard curve. This assay has a sensitivity of 8 pg/ml.

Measurement of TNF- Transcription Rate

TNF- and GAPDH transcription rates were measured using a modification of previously described nuclear run-on analysis (25). RAW cells were grown to approximately 80% confluence. LPS (100 ng/ml) was added, and the incubation was continued for the indicated duration. At each indicated time point, cells were lysed in 10 ml of lysis buffer, containing 10 mM Tris, pH 8.0, 2.5 mM MgCl₂, 0.25% Triton X-100, 0.3 M sucrose, 1 mM dithiothreitol, and nuclei were collected by centrifugation for 5 min at 500 × g. Isolated nuclei were manually counted, and equal numbers from each culture were incubated with 100 µCi of [-³²P]UTP (NEN Life Science Products) for 30 min at 37 °C. Plasmid containing GAPDH, full-length cDNA for TNF-, or the exon 4 TNF- cDNA fragment or plasmid without insert was alkaline-denatured and immobilized on nitrocellulose (Protran; Schleicher and Schuell) by vacuum filtration using a slot-blot manifold (Hoeffer) or a dot blot manifold (Bio-Rad). The membrane was rinsed with 6 × SSC, air-dried, and UV-cross-linked. Membranes were prehybridized overnight at 42 °C in 50% formamide, 5 × SSC, 2 × Denhardt's solution, 0.1% SDS, and 100 µg/ml denatured salmon testes DNA and then hybridized at 42 °C for 3 days in prehybridization buffer containing ³²P-labeled RNA extracted from the run-on reaction mixtures. Membranes were extensively washed (four washes with 2 × SSC, 0.1% SDS at room temperature for 5 min each and then 0.1 × SSC, 0.1% SDS at 60 °C for 30 min) and air-dried. The membranes were then analyzed with a PhosphorImager (Molecular Dynamics) before exposing to XAR film.

RAW cells were plated in six-well polystyrene tissue culture plates (Costar; Cambridge, MA) and allowed to reach 90% confluence. Monolayers were incubated with 67 µCi of [5,6-³H]uridine (NEN Life Science Products) in SFM for 10 min at 37 °C and were then treated with 5 µg/ml actinomycin D. At various times, aliquots of cells were washed with ice-cold SFM and lysed with SDS, and cell lysates were precipitated with ice-cold trichloroacetic acid. The trichloroacetic acid-insoluble precipitate was collected on glass fiber filters (GF/B, Whatman International, Ltd.; Maidstone, United Kingdom), air-dried, transferred to scintillation fluid (Ready Safe, Beckman Instruments, Inc.), and counted (LS5000TD, Beckman Instruments, Inc.).

RESULTS

Analysis of TNF- Transcript Size

To understand the anomaly in mRNA size and its role in TNF- gene expression, we first measured the expression pattern of TNF- in several macrophage cell lines as well as in primary cultured human monocyte-derived macrophages (Fig. 1A). In four mouse macrophage cell lines (lanes 1-8) and in primary cultured human macrophages (lanes 9 and 10), a discrete TNF- mRNA band was evident on Northern blots of RNA from adherent, unstimulated cells (lanes 1, 3, 5, 7, and 9). Stimulation with LPS (100 ng/ml) in serum-free medium for 1 h caused two effects in these cells: 1) an increase in TNF- mRNA band intensity, and 2) a decrease in electrophoretic mobility and broadening of the TNF- band on denaturing agarose/formaldehyde gels, indicating an increase in transcript size and heterogeneity after LPS stimulation (lanes 2, 4, 6, 8, 10). Similar changes occurred in cells incubated in serum-containing medium (data not shown). Higher resolution Northern analysis of RAW macrophage-like cell line TNF- mRNA on 3% agarose/formaldehyde gels (Fig. 1A, inset) indicated an approximately 200-240-nucleotide larger TNF- mRNA in the LPS-stimulated macrophages.

TNF- mRNA expression in unstimulated and LPS-stimulated macrophage cell lines and human peripheral blood monocyte-derived macrophages (hPBM).

Nuclear run-on analysis (Fig. 1B) showed TNF- was transcribed in unstimulated adherent RAW cells at 29% of the maximal rate achieved after LPS stimulation. As previously reported (15), elongation of TNF- transcripts is partially blocked in unstimulated macrophages, leading to 5 fragments of TNF- pre-mRNA. To examine the transcription rate of full-length TNF- mRNA, we also used a 3 cDNA fragment representing exon 4 to detect radiolabeled transcripts (Fig. 1B, hatched bars). The results of the run-on assays using the full-length TNF- cDNA and the exon 4 fragment were virtually identical.

LPS induced a 5-fold increases in TNF- mRNA levels 1 h post-stimulation (Fig. 1A). TNF- secretion, which was undetectable in unstimulated cells, increased to 4.87 ± 0.61 ng of TNF-/million cells 6 h after LPS treatment (Fig. 1C). These data suggest that LPS induces both TNF- mRNA stabilization and TNF- translation.

TNF- Transcript Size Heterogeneity Is Not Due to Differential Splicing in LPS-Stimulated Macrophages

Analysis of published TNF- cDNA and genomic sequences revealed that mature TNF- mRNA comprises 4 exons and 3 introns (20). To determine if one or both of the internal exons were excluded from the TNF- mRNA in unstimulated macrophages, random hexamer-primed reverse transcribed cDNA from adherent, unstimulated, and LPS (100 ng/ml for 1 h)-stimulated RAW cells were PCR-amplified (RT-PCR) using a primer pair that spanned the full length of the TNF- mRNA (Fig. 2A). Exclusion of an internal exon from unstimulated macrophage TNF- mRNA would result in amplification of a smaller RT-PCR product in these cells. In four experiments, only a single 1.6-kb product was amplified in unstimulated macrophages that was identical in size to the product in LPS-stimulated cells (Fig. 2B), thus excluding alternative TNF- splicing as a cause for the observed transcript heterogeneity.

Analysis of TNF- mRNA size shift.

Contribution of 3 Poly(A) Tail to TNF- mRNA Size Heterogeneity

Having excluded the possibility that a difference in internal TNF- sequence between unstimulated and LPS-stimulated macrophages could account for the heterogeneity, the contribution of heterogeneous TNF- transcript 3 poly(A) tail length to TNF- mRNA size heterogeneity was analyzed using an oligo(dT)/RNase H digestion assay (26) (Fig. 2C). The size of undigested (lanes 1, 3, 5, and 7) and digested (lanes 2, 4, 6, and 8) TNF- mRNA was determined by Northern blot analysis on 3% agarose/formaldehyde gels that had a resolution of approximately 20-30 nucleotides. Digesting TNF- mRNA from LPS-stimulated macrophages (lanes 4, 6, and 8) reduced its apparent size to that of TNF- mRNA in unstimulated cells (lane 1). By contrast, digesting TNF- mRNA from unstimulated cells did not cause any apparent change in TNF- transcript size (lanes 1 and 2). In comparison, GAPDH mRNA in unstimulated and LPS-stimulated macrophages were of similar size, and both were shortened equally by oligo(dT)/RNase H digestion. These results demonstrate that the TNF- mRNA 3 poly(A) tail is either absent or markedly truncated in unstimulated macrophages. Henceforth, we will refer to the short TNF- mRNA in unstimulated macrophages as hypoadenylated mRNA.

Stability of Hypoadenylated TNF- mRNA in Unstimulated Macrophages

The 3 poly(A) tail has been described as a cis acting transcript-stabilizing factor (27). Classically, 3 poly(A) tails shorten during translation. When its length is reduced below approximately 25 nucleotides, the minimal length required for poly(A) binding protein binding, degradation of the transcript body ensues (28). Taken together, the 3-4-fold increase in transcription rate and the 5-fold increase in steady state TNF- mRNA levels suggests that TNF- transcripts would be at least as stable in LPS-stimulated macrophages as in unstimulated cells. We measured the structural half-life of TNF- mRNA by blocking transcription with 5 µg/ml actinomycin D, which completely blocked [³H]uridine incorporation into RNA 10 min after treatment (Fig. 3A). The apparent half-life of the hypoadenylated TNF- mRNA in unstimulated TNF- mRNA was surprisingly long (30-49 min) (Fig. 3, B and C) compared with a 22-min half-life 1 h after LPS stimulation, which we have previously reported (25). These data indicate that a pool of relatively stable cytoplasmic hypoadenylated TNF- mRNA may be present in unstimulated macrophages. Of note, in the absence of LPS, the intensity of the TNF- mRNA band gradually decreased without a change in the apparent transcript length as would be expected if its degradation began with exonucleolytic shortening of the poly(A) tail (Fig. 3B, lanes 1-9). This transcript degradation pattern differs from the gradual shortening of TNF- mRNA that occurs after LPS stimulation (25).

Half-life of hypoadenylated TNF- mRNA in unstimulated RAW cells.

Polysome Association of TNF- mRNA in Unstimulated and LPS-stimulated Macrophages

The 3 poly(A) tail enhances recruitment of ribosomes to mRNA and increases translational efficiency (29). The polysome distribution of cytoplasmic TNF- mRNA was analyzed using sucrose density gradient centrifugation (Fig. 4). In LPS-stimulated RAW cells, 55% of the TNF- mRNA was associated with polysomes containing >5 ribosomes (Fig. 4, A and B, fractions 6-8). A minor component of predominantly hypoadenylated (poly(A)) TNF- mRNA was found in the ribonucleoprotein/monosome fractions (fractions 2 and 3) in the LPS-stimulated cells. On the contrary, virtually all of the TNF- mRNA in lysates from unstimulated cells was detectable in the ribonucleoprotein fractions (Fig. 4, C and D, fractions 2 and 3). No detectable TNF- mRNA was polysome-bound in these cells (lanes 6-8). These data show that the hypoadenylated TNF- mRNA in unstimulated macrophages is not associated with polysomes and is consistent with its translational dormancy as previously reported (17).

Polysome distribution of TNF- mRNA in unstimulated and LPS-stimulated RAW cells.

Effects of Translational Inhibitors on TNF- mRNA

It is possible the increased TNF- transcription induced by LPS overwhelms a limited capacity to remove 3 poly(A) tail. This possibility was addressed by analyzing TNF- mRNA content and size after treating RAW cells with cycloheximide, an agent that induces TNF- transcription (30) and increases TNF- mRNA level (31). Cycloheximide (CHX) increased TNF- transcription rate 80% (Fig. 5A), similar to the increases stimulated by LPS, and caused increased steady state TNF- mRNA levels, albeit less than with LPS. Interestingly, all of the TNF- mRNA present in cycloheximide-treated macrophages was the size of hypoadenylated TNF- mRNA in unstimulated cells (Fig. 5B, lanes 2-6), suggesting that a discrete signal provided by LPS and not by cycloheximide is needed to allow formation of polyadenylated TNF- mRNA. Alternatively, cycloheximide could directly block the capacity of the macrophage to accumulate polyadenylated TNF- mRNA. To test this possibility, RAW cells were first treated with 10 µg/ml cycloheximide for 3 h and then with LPS. At this concentration, cycloheximide inhibited protein synthesis >96% in these cells (data not shown). The cycloheximide-induced TNF- mRNA was the same size as hypoadenylated TNF- mRNA, but the TNF- transcripts lengthened by approximately 200 nucleotides after the addition of LPS to the cycloheximide-pretreated cells (Fig. 5C, lanes 2-4). Thus, the cycloheximide-treated cells were still capable of synthesizing fully polyadenylated TNF- mRNA in response to subsequent LPS stimulus. Furthermore, these data also suggest that regulation of TNF- poly(A) tail length can be dissociated from TNF- transcription and is independent of protein synthesis.

Effect of translational inhibitors on TNF- mRNA expression in RAW cells.

The loss of 3 poly(A) tails on TNF- transcripts in unstimulated macrophages may be the result (32), rather than the cause of their failure to associate with polysomes and initiate translation. If translational silencing and dissociation from polysomes were the primary event causing deadenylation of TNF- mRNA, then treatment with the polysome-disrupting agent puromycin would block the appearance of polyadenylated TNF- mRNA after LPS treatment. In fact, treating RAW cells with puromycin at concentrations (100 µg/ml) that blocked protein synthesis (data not shown) did not prevent the appearance of fully polyadenylated TNF- mRNA after LPS stimulation (Fig. 5D). These data demonstrate that translational inhibitors that preserve (cycloheximide) or disrupt (puromycin) polysome integrity fail to block accumulation of polyadenylated TNF- mRNA after stimulation with LPS. Thus, it is unlikely that the deadenylation of TNF- mRNA in unstimulated macrophages is caused by either ongoing translation or its dissociation from polysomes.

Lengthening of Cytoplasmic Hypoadenylated TNF- mRNA

Regulation of transcript poly(A) tail length is used in the developing oocyte and early embryo to orchestrate transcript translation (33). In these developmental systems, the truncated 3 poly(A) tail of translationally silent transcripts is lengthened prior to an increase in their translational efficiency (34-37). The effect of LPS stimulation on the size of preexisting cytoplasmic TNF- mRNA was analyzed in RAW cells treated with 5 µg/ml actinomycin D to block new TNF- mRNA synthesis (Fig. 6, lanes 6-10). Blocking transcription unmasked a LPS-induced lengthening of the cytoplasmic pool of hypoadenylated TNF- mRNA (lanes 8-10). The TNF- transcripts reached a maximal size that was intermediate between hypoadenylated and fully adenylated TNF- mRNA. The TNF- mRNA levels in these cells were not sufficient for polysome analysis or oligo(dT)/RNase H digestion, but these results suggest that LPS may stimulate the readenylation of preexisting cytoplasmic pools of TNF- mRNA. Despite >97% inhibition of transcription in RAW cells treated with 5 µg/ml actinomycin D for 10 min, these cells were still capable of limited TNF- secretion (Table I).

Effect of LPS stimulation on apparent transcript length of existing pools of TNF- mRNA in RAW cells.

Table I. Transcription-independent TNF- secretion LPS (100 ng/ml)-induced TNF- secretion was measured by enzyme-linked immunosorbent assay in RAW cells with or without 10-min pretreatment with 5 µg/ml actinomycin D. Mean ± S.E. values of three experiments are shown. TNF- protein in culture supernatants 0 ha 0.5 h 1 h 2 h 6 h pg/ml No pretreatment 63.1  ± 4.4 354  ± 24b 3269  ± 771c 10,119  ± 666c 9839  ± 827c Actinomycin D 179  ± 7.9 334  ± 66b 730  ± 55c 748  ± 46c 443  ± 44b a Time post-LPS stimulation. b p < 0.05 compared with time 0. c p < 0.003 compared with time 0.

DISCUSSION

Macrophage activation causes dramatic changes in posttranscriptional processing of TNF- mRNA (16, 17). As previously reported in other macrophage models (11, 16, 17), basal TNF- protein secretion was not detectable in adherent RAW macrophages (Fig. 1C) in the face of active TNF- transcription (Fig. 1B) and the accumulation of cytoplasmic TNF- mRNA (Fig. 1A). We also could not detect TNF- protein in unstimulated RAW cell lysates (data not shown). Han et al. (17) directly showed that translation of TNF- was blocked in unstimulated RAW cells and LPS stimulation reversed the inhibition of TNF- protein synthesis. We have extended these data by showing that virtually all detectable TNF- mRNA in unstimulated RAW macrophages was not associated with polysomes (Fig. 4), indicating TNF- translation initiation was blocked in these cells (38). LPS treatment induced both the appearance of polysome-associated TNF- mRNA (Fig. 4) and onset of TNF- secretion (Fig. 1C).

We previously reported preliminary evidence suggesting TNF- transcripts are shorter in unstimulated adherent RAW macrophages than in LPS-stimulated cells (18) and now report similar findings in human monocyte-derived macrophages and three other mouse macrophage cell lines (Fig. 1). Ulich et al. (19) reported similar TNF- mRNA size heterogeneity in rat alveolar macrophages in vitro and in rat spleen and liver in vivo, but the mechanisms by which these shifts in apparent TNF- mRNA length occur were not described. In each of these cases TNF- protein was not detectable in the absence of longer TNF- transcripts. In the present study, enzymatic removal of the 3 poly(A) tail reduced the apparent length of TNF- mRNA in LPS-stimulated cells by approximately 200 nucleotides. The same digestion of RNA from unstimulated cells did not cause any detectable change in the apparent 1.6-kb TNF- mRNA (Fig. 2C). Oligo(dT)/RNase H digestion shortens poly(A) tails to within 20 nucleotides of the transcript body (35), and the resolution of our electrophoresis system was 20-30 nucleotides. Thus, these data indicate that the TNF- mRNA 3 poly(A) tail is <30 nucleotides long in unstimulated macrophages. Our inability to amplify TNF- from unstimulated RAW cells using oligo(dT)_12-18-primed RT-PCR (data not shown) suggests the TNF- poly(A) tail may be even shorter in these cells. In contrast, poly(A) tail length of the constitutively expressed housekeeping gene transcript GAPDH was unaffected by the macrophage activation state (Fig. 2C). A similar analysis of other cytokine transcripts, including granulocyte/macrophage colony-stimulating factor, interleukin-6, and interleukin-1, was not possible in this study because these transcripts were not detectable in unstimulated RAW cells (data not shown).

In unstimulated RAW cells, the apparent TNF- transcription rate measured by nuclear run-on assay was almost one-third the maximal rate achieved in LPS-stimulated cells (Fig. 1B). Biragyn and Nedospasov (15) recently reported that elongation of TNF- transcripts is partially blocked in unstimulated macrophages despite modest rates of transcription initiation. These authors suggested that this process may limit effective transcription in unstimulated macrophages. Because the TNF- gene contains a single poly(A) signal site near its 3 terminus, incomplete transcription could account for the lack of 3 polyadenylation. Three lines of evidence suggest that this is not the major mechanism responsible for the accumulation of hypoadenylated TNF- mRNA in unstimulated macrophages. First, Northern analysis of cytoplasmic RNA from unstimulated RAW cells only showed a single 1.6-kb TNF- band. Second, RT-PCR amplification of cytoplasmic RNA using a primer complementary to the TNF- 3 terminus generated an amplicon of the expected size in unstimulated macrophages. Finally, the nuclear run-on assay that demonstrated basal TNF- transcription utilized an exon 4 TNF- cDNA fragment to detect radiolabeled transcripts (Fig. 1B).

Detection of only 1.6-kb TNF- mRNA in unstimulated macrophages, in the face of ongoing TNF- transcription, suggests either a block in nuclear 3 polyadenylation of TNF- mRNA or rapid TNF- transcript deadenylation in these cells. Specific cytoplasmic deadenylation of polyadenylated mRNA has been described in mammalian cells (39, 40). Poly(A) tails of mRNA are nearly ubiquitous in higher eukaryotes. The only mRNAs known to lack poly(A) are those encoding the major histones that possess unique 3 secondary structure that is not present in TNF- mRNA (41). To our knowledge, synthesis and nuclear export of both adenylated and deadenylated forms of the same transcript have not been reported. Therefore, we believe that TNF- mRNA is synthesized and exported to the cytoplasm with a 3 poly(A) tail, which is rapidly removed in the cytoplasm of unstimulated macrophages.

The process responsible for the accumulation of deadenylated TNF- mRNA in unstimulated RAW cells appears to be distinct from the gradual poly(A) tail shortening of polyadenylated TNF- mRNA in LPS-stimulated cells. Gradual TNF- mRNA poly(A) tail shortening is evident within 30 min of LPS stimulation (Fig. 2C) and proceeds over the next 8 h in RAW cells as we have previously described (25), indicating a typical distributive, translation-dependent deadenylation process (42). By comparison, only hypoadenylated, ribonucleoprotein-associated TNF- mRNA was detectable in unstimulated macrophages (Figs. 1, 2C, and 4).

Surprisingly, hypoadenylated TNF- mRNA in unstimulated RAW cells was more stable (Fig. 3C) than polyadenylated TNF- mRNA in LPS-stimulated cells (25). Generally, mRNA deadenylation is followed by 5 cap removal and transcript degradation (43). The relative stability of the hypoadenylated TNF- mRNA pool may be due to its translational dormancy, since at least some RNases are polysome-associated (44). By failing to associate with polysomes, hypoadenylated TNF- mRNA in unstimulated macrophages may not be accessible to these RNases.

The data presented in this report suggest that unstimulated macrophages transcribe TNF- but convert TNF- mRNA to a relatively stable, hypoadenylated, translationally dormant form. The possible importance of this process as a negative regulator of TNF- expression in unstimulated macrophages is apparent. The advantage to the host of this ostensibly inefficient process of synthesizing translationally dormant TNF- mRNA is less obvious, but its presence suggests that the pool of hypoadenylated TNF- mRNA may serve an important function.

Gene regulation in developmental systems may provide clues about the role of TNF- mRNA deadenylation in macrophages. Translational activation of certain maternal genes during oocyte maturation, fertilization, and early embryogenesis is orchestrated by cytoplasmic deadenylation and poly(A) elongation (33). For certain genes, premature expression is prevented by truncating the poly(A) tails of newly synthesized transcripts before translation is initiated (45). Stable pools of these translationally dormant, hypoadenylated transcripts are stored for eventual expression. Translation is subsequently initiated by poly(A) tail elongation and continues until the 3 poly(A) tail is removed. In many ways TNF- mRNA in macrophages behaves like the developmental mRNAs that are regulated by cytoplasmic polyadenylation. Unstimulated macrophages contain only hypoadenylated cytoplasmic TNF- mRNA, which is relatively stable and translationally silent (17) (Figs. 1, 2, 3, 4). Moreover, in macrophages pretreated with actinomycin D to block synthesis of new TNF- mRNA, LPS stimulated an elongation of TNF- mRNA (Fig. 6) and low level secretion of TNF- protein (Table I), suggesting TNF- mRNA may undergo polyadenylation and translation in the macrophage cytoplasm. This process may provide the macrophage with a means of rapid, limited TNF- synthesis. Furthermore, this alternative pathway may allow TNF- synthesis to occur in the face of transcriptional arrest. Cytoplasmic deadenylation/polyadenylation in developmental systems requires the canonical nuclear polyadenylation signal, AAUAAA, and a second, less defined U-rich 3-untranslated region element that is similar to several stretches of AU-rich sequences in the TNF- 3-untranslated region (46).

We have identified two possible new regulatory sites in TNF- expression, 3 poly(A) tail metabolism and association of TNF- mRNA with polysomes. Additionally, macrophages may be able to readenylate cytoplasmic pools of deadenylated TNF- mRNA, which may provide an alternative, transcription-independent source of translationally active TNF- mRNA (Fig. 7).

Proposed mechanism of 3 polyadenylation in regulating TNF- expression.