Introduction
mRNAs associate with proteins to form messenger ribonucleoprotein particles (mRNPs). Microarray analyses of mRNPs from mammalian cells have revealed that mRNA-binding proteins associate with unique subsets of mRNAs to coordinately regulate their localization, translation and/or degradation during disparate biological processes1, 2. mRNP composition can change with cellular growth states, and mRNA species are often found in multiple mRNP complexes. These mRNA subsets represent post-transcriptional operons that integrate production of protein groups needed collectively for a biological process. Coordinate expression of a given mRNA subset is facilitated by related 'untranslated sequence elements for regulation' (USERs)2.
There are many types of USERs. The best characterized is a family of sequences known as AREs. AREs target degradation or translational suppression of many mRNAs encoding oncoproteins, cytokines, cell-cycle regulators and signaling proteins3, 4, 5. This is achieved through the association of AREs with one or more ARE-binding proteins and in some cases microRNAs5, 6. The AUF1/hnRNP D family of ARE-binding proteins promotes degradation of mRNAs encoding cytokines such as IL-3 and GM-CSF and cell-cycle regulators such as p16INK4a, p21WAF1/CIP1 and cyclin D1 (refs. 7–9). AUF1 is comprised of four isoforms, of 37, 40, 42 and 45 kDa, generated by alternative precursor mRNA (pre-mRNA) splicing10, 11, 12. They shuttle between the nucleus and cytoplasm13, 14 and associate with heat-shock proteins hsc70-hsp70, translation initiation factor eIF4G and poly(A)-binding protein (PABP) to promote mRNA degradation. mRNA degradation requires dissociation of AUF1 from eIF4G and ubiquitin-dependent destruction of AUF1 by proteasomes15, 16.
Estimates are that mRNAs bearing AREs represent as much as 5%–8% of the human transcriptome17. Because individual ARE-binding proteins associate with selected subsets of ARE-mRNAs, we initiated experiments to define the subset of mRNAs associated with AUF1 in cells of human origin. We first immunopurified AUF1-associated mRNPs from cytoplasmic lysates, then used microarray analyses to identify the purified mRNA subset. One AUF1-associated target was mRNA encoded by the MYC proto-oncogene (formerly known as c-MYC). Proper control of the MYC gene is essential, as its deregulation is central to formation of most tumors18, 19. Knockdown of AUF1 using RNA interference (RNAi) had no discernable effect upon MYC mRNA abundance. We thus examined effects on translation and found that AUF1 knockdown reduced translation of MYC mRNA in an ARE-dependent fashion. AUF1 knockdown reduced cell proliferation as well. This led us to initiate experiments to identify the mechanisms by which AUF1 promotes translation of MYC mRNA and determine whether AUF1 affects cellular proliferation via a MYC-dependent pathway.
Results
AUF1 controls MYC expression
Our initial goal was to define the subset of mRNAs associated with AUF1 using the human chronic myelogenous leukemia cell line K562 as a model20. mRNPs were immunoprecipitated from cytoplasmic lysates using antibody specific for AUF1, or preimmune serum as a control. Precipitated mRNA was purified, labeled and used to interrogate microarrays for their identification. One AUF1-associated target was mRNA encoded by the MYC gene (data not shown). This was verified by quantitative, real-time reverse-transcription PCR (qRT-PCR) of MYC mRNA using immunoprecipitates (Supplementary Fig. 1 online). The remainder of our studies focused on this gene product, for three reasons. (i) MYC encodes a transcription factor often central to tumorigenesis and aberrantly expressed in many human tumors18, 19. (ii) MYC contains an ARE in its 3' untranslated region (UTR) that binds AUF1 with high affinity in vitro, consistent with the mRNP immunoprecipitation results10, 21, 22. (iii) Cell-free mRNA decay experiments suggested that partially purified AUF1 could promote degradation of MYC mRNA23. Together, these observations indicate that AUF1 contributes to MYC expression.
To examine control of MYC expression by AUF1 in cells, AUF1 abundance was reduced using RNAi and the resulting effects on MYC mRNA levels were assessed by qRT-PCR. K562 cells were transfected with a plasmid encoding either a short hairpin RNA (shRNA) directed against AUF1 to reduce its expression, or an shRNA of random sequence as a control. After three days, the AUF1 abundance was reduced 90% compared with the control shRNA transfection (Fig. 1a,b). Contrary to our expectation that MYC mRNA abundance would increase, it was unchanged both in total RNA and in cytoplasmic RNA (Fig. 1c). Similar results were obtained upon AUF1 knockdown in other cell types, including HeLa (human cervical carcinoma; Supplementary Fig. 2a–c online), HT-29 (human colon carcinoma; data not shown) and THP-1 (human promonocytic leukemia; data not shown). In contrast to the lack of effect on MYC mRNA levels, AUF1 knockdown led to three-fold reduction of cytoplasmic and total cellular MYC protein, compared with cells transfected with the control shRNA plasmid (Fig. 1d,e). Similar decreases in MYC abundance were observed upon knockdown of AUF1 in other cell types, including HeLa (Supplementary Fig. 2d,e), HT-29 (data not shown) and THP-1 (data not shown). Reduction in MYC levels upon AUF1 knockdown was confirmed using two small interfering RNAs (siRNAs) directed against different sequences within the AUF1 mRNA (data not shown); this result argues against off-target effects of the shRNA. As reduction of AUF1 had no significant effect (P > 0.05) upon either cytoplasmic or total cellular MYC mRNA levels (Fig. 1c), we conclude that AUF1 knockdown decreases translation of MYC mRNA. Thus, the normal function of AUF1 may be to promote translation of MYC mRNA.
Figure 1: Effects of AUF1 abundance on MYC expression.
Cells expressing negative control shRNA or shRNA targeting AUF1 were examined for AUF1 protein and MYC mRNA and protein abundances in cytoplasmic or whole-cell lysates. (a) Representative AUF1 and 
-tubulin western blots using whole-cell lysates. (b) Western blot bands quantified and normalized to -tubulin. AUF1 levels were reduced 90% (two-tailed t-test: n = 6, P = 0.0003). (c) MYC mRNA levels were assessed in cytoplasmic or whole-cell RNA by qRT-PCR. AUF1 knockdown had no significant effect upon MYC mRNA levels (two-tailed t-test: n = 6, P = 0.7244 for cytoplasmic RNA and 0.6629 for whole-cell RNA). (d) Representative western blots using antibodies for MYC and -tubulin. (e) Western blots as in d were quantified. AUF1 knockdown appreciably reduced cellular MYC levels compared with control shRNAs (two-tailed t-test: n = 6; P = 0.002 for cytoplasmic lysate and 0.003 for whole-cell lysate). (f,g) Plasmids encoding single AUF1 isoforms24 were transfected individually or combined (Com) into K562 cells. Empty plasmid was a control (Con). f shows representative western blot analysis of cell lysates using antibodies to AUF1, MYC and -tubulin. g shows analyses of MYC mRNA abundance after AUF1 overexpression. There was no effect of AUF1 on MYC mRNA levels (ANOVA: n = 5, P = 0.7011). In this and following figures, data are represented as mean 
s.d.; **P < 0.01.
A prediction of this hypothesis is that overexpression of AUF1 may promote translation of MYC mRNA. K562 cells were transfected either with a pcDNA3 plasmid encoding one AUF1 isoform24 or with all four plasmids combined (Com). Empty plasmid pcDNA3 served as a control (Con). AUF1 increased 65% (for p45) to 300% (for p40, p42 or all four combined) after transfection of the plasmids; increases in p37 could not be ascertained, as it was not readily detectable in the control (Fig. 1f). Nonetheless, overexpression of either individual or all four isoforms elevated MYC levels three- to four-fold (Fig. 1f) without affecting MYC mRNA levels (Fig. 1g). Together, AUF1 knockdown and overexpression experiments strongly suggest that AUF1 promotes translation of MYC mRNA.
AUF1 controls translation via the MYC ARE
To confirm this conclusion and to ascertain whether the translational effect was exerted through the ARE, we constructed matched sets of firefly luciferase reporter genes containing the MYC ARE in the sense or antisense orientation within the 3' UTR. These were transiently transfected into K562 cells together with a plasmid expressing either the control shRNA or the AUF1 shRNA. A plasmid expressing Renilla luciferase was cotransfected as an internal control for normalization of firefly luciferase. Reduction of AUF1 had no effect upon levels of the reporter ARE mRNA in either total cellular RNA or cytoplasmic RNA (Fig. 2a). By contrast, knockdown of AUF1 reduced firefly luciferase activity from the reporter ARE mRNA by 
60% but did not significantly (P > 0.05) affect luciferase activity from the reporter mRNA with the ARE in the antisense orientation (Fig. 2b*). Similar results were obtained upon knockdown of AUF1 in HeLa cells (Supplementary Fig. 3 online). Overexpression of individual AUF1 isoforms or all four combined increased luciferase activity in an ARE-dependent fashion (data not shown), consistent with the effects of AUF1 overexpression upon endogenous MYC translation (see Fig. 1). Finally, to determine whether AUF1 associates specifically with the reporter ARE-mRNA in cells, lysates were prepared from K562 cells transfected with the luciferase reporter plasmids; luciferase plasmid lacking any additional sequences (pGL3) served as a further control. These cells did not express either the control or AUF1 shRNA. mRNP immunoprecipitation with AUF1 antibody and qRT-PCR of firefly luciferase mRNA revealed luciferase–sense ARE mRNA (Luc/ARE-s), but 500-fold less (s.d. = 70, n = 3) luciferase–antisense ARE mRNA (Luc/ARE-as); luciferase mRNA lacking any additional sequence (pGL3) was undetectable (Fig. 2c,d).
Figure 2: AUF1 affects translation via the MYC ARE.
K562 cells were cotransfected with plasmids expressing either control or AUF1 shRNA; plasmid pRL-SV40 encoding Renilla luciferase; and either pGL3-Luc-MYC ARE-s or pGL3-Luc-MYC ARE-as plasmids. (a) AUF1 knockdown did not affect luciferase–MYC ARE reporter mRNA abundance (two-tailed t-test: n = 6, P = 0.6199 and 0.7734 for cytoplasmic and whole-cell RNA, respectively). (b) AUF1 knockdown decreased luciferase activity from the luc–MYC ARE reporter compared with the luc-MYC ARE-as reporter plasmid (two-tailed t-test: n = 6, P = 0.7495 for control shRNA; 0.0005 for AUF1 shRNA). (c,d) Transfected cells were analyzed by mRNP immunoprecipitation using anti-AUF1 and qRT-PCR. AUF1 preferentially associates with luciferase mRNA containing the MYC ARE in the sense orientation. c shows representative qRT-PCR amplification plot. 
Rn, fluorescence (see Methods). d shows representative agarose gel analysis of PCR end products from c. Tracks 1–4 show 100-bp DNA markers and qRT-PCR products from immunoprecipitates of cells transfected with pGL3, pGL3-Luc-MYC ARE (antisense) and pGL3-Luc-MYC ARE, respectively. The PCR product was consistent with the predicted 189-bp length. Luc, firefly luciferase; s, sense; as, antisense. Error bars represent s.d.
Collectively, the reporter gene experiments indicate that AUF1 promotes translation of the reporter ARE mRNA without affecting mRNA levels. These observations are consistent with those for endogenous MYC mRNA (see Fig. 1). Moreover, the ARE dependence of the interaction of AUF1 with the reporter mRNA suggests that AUF1 associates with MYC mRNA via its interaction with the MYC ARE. However, this does not exclude the possibility that AUF1 associates with MYC mRNA via interactions within additional regions of the mRNA as well. Nonetheless, the AUF1-ARE association can confer translational regulation.
AUF1 knockdown redistributes MYC mRNA in polyribosomes
The luciferase reporter experiments also reduced the possibility that AUF1 affects MYC protein stability, thus bolstering the conclusion that AUF1 affects translation of MYC mRNA. To further substantiate effects of AUF1 on MYC translation, we performed polyribosome-profile experiments using cytoplasmic lysates of K562 cells. The majority of AUF1 cosedimented with free RNP (Fig. 3a,b, fractions 1 and 2). However, a portion of AUF1 cosedimented with the 40S and 60S ribosomal subunits (fractions 3 and 4), consistent with earlier results25. In cells transfected with control siRNA, 75% of MYC mRNA cosedimented with about one to three ribosomes (Fig. 3c, fractions 5–7, filled bars). Upon knockdown of AUF1, almost 60% of MYC mRNA cosedimented with ribosomal subunits (Fig. 3c, fractions 3–5, open bars). AUF1 knockdown had little effect upon either the distribution of GAPDH mRNA (Fig. 3d) or the bulk distribution of polyribosomes (compare Fig. 3e,f). By contrast, AUF1 overexpression increased the proportion of MYC mRNA in polyribosomal fractions (Fig. 3g) without affecting the distribution of GAPDH mRNA (Fig. 3h). Together, these results indicate specific translational control of MYC mRNA by AUF1.
Figure 3: AUF1 abundance affects polyribosome distribution of MYC mRNA.
Cytoplasmic lysates of untransfected K562 cells or cells transfected with control siRNA or AUF1 siRNA were subjected to polyribosome profile analyses. (a) A representative polyribosomal profile from untransfected K562 cells. The 254-nm trace obtained during collection of fractions is shown, with positions of free RNP, 40S and 60S ribosomal subunits, 80S ribosomes and polyribosomes indicated at the top. (b) Relative distributions of AUF1 and -tubulin were obtained by western blot analyses of the gradient shown in a. Fractions: 1 and 2, unbound RNPs (Unb); 3–5, 40S, 60S, 80S (Mono); 6–11, polyribosomes. (c,d) Relative distributions of MYC (c) and GAPDH mRNAs (d) in polyribosome gradients after AUF1 knockdown. Data shown are mean s.d. from three independent experiments. Ribosome subunit distributions are as described in b. Control, control siRNA–transfected cells; AUF1, AUF1 siRNA–transfected cells. (e,f) Representative polyribosomal profiles from K562 cells transfected with control siRNA (e) or AUF1 siRNA (f). (g,h) Relative distributions of MYC (g) and GAPDH mRNAs (h) in polyribosome gradients after overexpression of four AUF1 isoforms combined. Data shown are means s.d. from three independent experiments. Ribosome subunit distributions are as described in b. Control, control pcDNA3–transfected cells; AUF1, cells transfected with plasmids expressing all AUF1 isoforms. Sedimentation for all gradients is from left to right.
TIAR inhibits translation of MYC mRNA via the ARE
To explore the mechanism(s) by which AUF1 promotes translation of MYC mRNA, we investigated interaction of MYC mRNA with T-cell internal antigen-1 (TIA-1)–related protein (TIAR), an ARE-binding protein and MYC translational suppressor26, 27. First, our mRNP immunoprecipitation experiments showed that TIAR antibody specifically precipitated MYC mRNA (Fig. 4a, upper gel, lane 5; see also Fig. 4b) and luciferase–MYC ARE reporter mRNA from K562 cytoplasmic lysates (Supplementary Fig. 4 online), indicating ARE-dependent association of TIAR and MYC mRNA in vivo. Knockdown of TIAR using siRNA increased MYC protein abundance approximately two-fold (Fig. 4c,d) without affecting MYC mRNA abundance (Supplementary Fig. 5 online), consistent with other cell types27. Likewise, TIAR knockdown elevated luciferase activity approximately two-fold in MYC ARE reporter gene experiments (Supplementary Fig. 6 online). Thus, TIAR suppresses translation of MYC mRNA in an ARE-dependent manner.
Figure 4: Effects of AUF1 and TIAR on MYC expression and mRNA association.
(a,b) K562 cells were transfected with control, AUF1 or TIAR siRNA together with plasmid pGL3-Luc-MYC ARE. Cytoplasmic lysates were used for mRNP immunoprecipitation (IP) assays with AUF1 or TIAR antibodies and MYC and luciferase mRNA abundances examined in immunoprecipitates. (a) Representative agarose gels of qRT-PCR end products for MYC (upper gel) and luciferase (Luc, lower gel) mRNAs. (b) qRT-PCR measurements of MYC mRNA abundances in immunoprecipitates. Abundances after control siRNA transfections were set to 1. (Two-tailed t-test: n = 5, P = 0.003 and 0.009 for TIAR and AUF1 siRNA transfections, respectively). (c,d) K562 cells were treated with control, AUF1, TIAR, or both AUF1 and TIAR siRNAs. c shows representative western blot analyses of whole-cell lysates using antibodies to MYC and -tubulin. d shows quantitative analyses of MYC and -tubulin western blots. (ANOVA: n = 5, P = 0.0001. Dunnett post-test for data from cytoplasmic lysate: control versus AUF1 siRNA transfection, P = 0.0001; control versus TIAR siRNA transfection, P = 0.0052; control versus AUF1 and TIAR siRNA cotransfection, P = 0.0858. Dunnett test of data from whole-cell lysate showed similar P values.) Error bars represent s.d.
AUF1 and TIAR compete for binding to MYC ARE
As both AUF1 and TIAR bind the MYC ARE, with opposite effects on translation (see Fig. 4c,d), we investigated their contributions to MYC expression by examining their relative interactions with MYC mRNA. TIAR abundance was unaffected by AUF1 knockdown and vice versa (Supplementary Fig. 7a–c online). Accordingly, we examined their competitive binding to MYC mRNA. After AUF1 knockdown by siRNA transfection, TIAR antibody precipitated three-fold more MYC mRNA compared with the control siRNA transfection (Fig. 4a, upper gel, compare lane 6 with lane 5; see also Fig. 4b). Immunoprecipitates contained equal amounts of precipitated TIAR protein, however (Supplementary Fig. 7d). Thus, differences in immunopurified MYC mRNA were not attributable to accessibility of antibodies to TIAR. Likewise, after TIAR knockdown by siRNA transfection, AUF1 antibody precipitated almost three-fold more MYC mRNA (Fig. 4a, upper gel, compare lane 3 with lane 2; see also Fig. 4b), but similar amounts of AUF1 protein, compared with the control siRNA transfection (Supplementary Fig. 7d). mRNP immunoprecipitation assays with K562 cells transfected with a luciferase–MYC ARE reporter gene yielded results comparable to those for MYC mRNA (Fig. 4a, compare upper and lower gels). Together, these results suggest that translational inhibition of MYC mRNA by AUF1 knockdown results from association of more TIAR with MYC mRNA. Consistent with this interpretation, overexpression of TIAR reduced the amount of MYC mRNA associated with AUF1 (Fig. 5a, upper gel, compare lane 3 with lane 2; see also Fig. 5b). Likewise, overexpression of AUF1 reduced the amount of MYC mRNA associated with TIAR (Fig. 5a, upper gel, compare lane 6 with lane 5; see also Fig. 5b). Similar results were obtained in experiments using the luciferase–MYC ARE reporter gene (Fig. 5a, compare upper and lower gels). Thus, overexpression of one protein favors its association with either endogenous MYC mRNA or the luciferase ARE reporter mRNA, at the expense of the other protein. In sum, these data indicate that AUF1 and TIAR competitively bind the MYC ARE in vivo and suggest that a dynamic ratio of AUF1 and TIAR association with MYC mRNA could affect cellular MYC levels post-transcriptionally. Moreover, these data indicate that AUF1 may indirectly promote translation by reducing association of TIAR, a translation suppressor26, 27, with MYC mRNA.
Figure 5: Effects of AUF1 or TIAR overexpression on ARE association.
K562 cells were transfected with empty expression plasmid (control) or plasmids expressing AUF1 (all four isoforms together) or TIAR (pMT2-HA-TIAR), together with plasmid pGL3-Luc-MYC ARE. Cytoplasmic lysates were used for mRNP immunoprecipitation (IP) assays with antibodies to AUF1 or TIAR followed by qRT-PCR to measure MYC and luciferase–MYC ARE mRNA abundances in immunoprecipitates. (a) Representative agarose gel analyses of qRT-PCR end products for MYC (upper gel) and luciferase-MYC ARE (Luc, lower gel) mRNAs. (b) qRT-PCR showing MYC mRNA abundances in immunoprecipitates (two-tailed t-test: n = 5, P = 0.0002 and 0.0005 for TIAR and AUF1 plasmid transfections, respectively). Error bars represent s.d.
Full size image (23 KB)AUF1 also directly promotes translation of MYC mRNA
As noted earlier (see Fig. 4c,d), TIAR knockdown increased MYC levels. However, knockdown of AUF1 together with TIAR reversed this effect of TIAR knockdown on MYC levels (Fig. 4c,d) without affecting MYC mRNA levels (Supplementary Fig. 5c). If AUF1 promotes translation of MYC mRNA simply by inhibiting suppression by TIAR, then MYC levels would be expected to remain elevated after combined TIAR and AUF1 knockdown. Thus, the double-knockdown result suggests that AUF1 may directly activate MYC translation in addition to blocking association of the TIAR suppressor protein with MYC mRNA. This direct mechanism might involve AUF1 interaction with the eIF4G–PABP complex. These protein interactions occur in cells and in vitro15, 28 (see Discussion).
Control of cell proliferation by AUF1 is MYC dependent
In addition to its effects on MYC protein levels, knockdown of AUF1 altered expression of a number of cell-cycle regulators at the mRNA level (data not shown). Indeed, AUF1 knockdown decreased proliferation by 65% for K562 cells (Fig. 6a), 63% for HeLa cells (Supplementary Fig. 8 online) and 60% for both HT-29 and THP-1 cells (data not shown). These results are consistent with the observation that overexpression of AUF1 promotes proliferation of human WI-38 diploid fibroblasts9. By contrast, TIAR knockdown, which increased MYC levels, increased cell proliferation more than two-fold (Supplementary Fig. 9 online). Cell proliferation returned to normal after knockdown of both AUF1 and TIAR (Supplementary Fig. 9). These data indicate that the two RNA-binding proteins have opposite effects on proliferation, and as both control MYC expression, both might regulate proliferation at least in part through a MYC-dependent pathway. Consistent with this hypothesis is the observation that inducible overexpression of MYC in a background of AUF1 knockdown in K562 cells reversed the proliferation defect arising from AUF1 knockdown (Fig. 6b–d). In addition, knockdown of both MYC and TIAR reduced proliferation, thus reversing the proliferation-inducing effects observed with TIAR knockdown alone (Supplementary Fig. 10 online). We thus conclude that both AUF1 and TIAR control cell proliferation by regulating MYC translation. Although we cannot exclude participation by other pathways, a MYC-dependent pathway is a major effector of the AUF1-TIAR regulatory axis.
Figure 6: AUF1 controls cellular proliferation.
(a) AUF1 knockdown inhibited cell proliferation 68% compared with the control siRNA transfection (two-tailed t-test: n = 10, P = 0.0009). Error bars represent s.d. (b–d) Zn2+-inducible overexpression of MYC (see Methods) reverses the cellular-proliferation defect conferred by AUF1 knockdown. (b) Representative western blot analyses. (c) Quantitative analyses of MYC and AUF1 from western blots. Here and in d, % of control refers to control siRNA transfection for uninduced MYC. (ANOVA: n = 6, P = 0.0001. Dunnett post-test for AUF1 protein data: control (-) versus control (+), P = 0.5734; control (-) versus AUF1 siRNA (+), P = 0.0001; control (-) versus AUF1 siRNA (-), P = 0.0001. Dunnett test for MYC data: control (-) versus control (+), P = 0.0001; control (-) versus AUF1 siRNA (+), P = 0.0582; control (-) versus AUF1 siRNA (-), P = 0.0004). (d) Measurements of cell proliferation by MTS assay. (ANOVA: n = 10, P = 0.0001. Dunnett post-test: control (-;) versus control (+), P = 0.0003; control (-) versus AUF1 siRNA (+), P = 0.0032; control (-) versus AUF1 siRNA (-), P = 0.0001). Error bars represent s.d.
Full size image (36 KB)Discussion
Transcriptional control of MYC is well documented. However, a plethora of multilayered, post-transcriptional control systems regulate MYC as well. These control systems include nucleocytoplasmic export of the mRNA by nuclear eIF4E29; cap-dependent translation initiation at alternative sites to generate three distinct protein isoforms30, 31; cap-independent translation directed by a 5' UTR internal ribosome entry site (IRES)32; translational repression by TIAR and cytoplasmic polyadenylation element–binding protein (CPEB)27, 33; translational activation by AUF1 (this work); mRNA degradation mediated by the 3' UTR ARE and a 182-nucleotide segment within the 3' coding region34, 35, 36; and phosphorylation, ubiquitination and acetylation of MYC37.
Earlier work had suggested that AUF1, partially purified from cytoplasmic extracts, could promote MYC mRNA decay in a cell-free system23. Unexpectedly, we found that knockdown of all AUF1 isoforms had no effect upon MYC mRNA abundance in cells, but instead regulated translation of MYC mRNA. However, AUF1 knockdown experiments have established its participation in degradation of other ARE-bearing mRNAs, such as those encoding p16INK4a, p21WAF1/CIP1, growth arrest– and DNA damage–inducible gene 45 (GADD45), TNF and IL-1
(refs. 8,9,38). Thus, AUF1 does participate in degradation of at least a subset of ARE mRNAs. Nonetheless, we do not know the exact reasons for the discrepancy. Two possibilities are as follows. (i) Simultaneous knockdown of all four AUF1 isoforms has been found to have no effect upon expression of an IL-3 ARE reporter mRNA or GM-CSF; only selective knockdown of p40 and p45 AUF1 isoforms leads to ARE mRNA stabilization7. Thus, at least for some ARE mRNAs, the balance of AUF1 isoform expression seems to be essential for mRNA degradation. In regard to this observation, it has been argued that knockdown of all four AUF1 isoforms might obscure its destabilizing effects, as different isoforms with opposing functions may regulate some ARE mRNAs38. (ii) Alternatively, MYC mRNA contains a coding-region instability determinant, the CRD, which may function to promote MYC mRNA degradation in vivo even in an AUF1-depleted background35, 36. Consistent with this possibility is the observation that homozygous deletion of the MYC 3' UTR in transgenic mice did not alter steady-state levels of the mRNA in the four tissues examined (spleen, thymus, liver and kidney)39. However, additional experiments will be required to fully address the coding-region hypothesis for other cell types.
As noted above, AUF1 regulates stability of numerous ARE mRNAs. As AUF1 binds several hundred target transcripts (ref. 8, and B.L. and G.B., unpublished data), it is likely that AUF1 controls translation of other ARE mRNAs as well. Our future work will identify and characterize this mRNA subset. However, to our knowledge, the work presented here is the first to directly demonstrate translational regulatory effects of AUF1. In particular, our results uncover a novel mode of MYC regulation such that the balance of AUF1 and TIAR associated with the MYC ARE controls mRNA translation rate. Thus, AUF1 may be essential to maintain MYC expression at appropriate levels. Consistent with this hypothesis, AUF1 and MYC mRNA share similar spatio-temporal expression patterns during development and in adult tissues40. By contrast, TIAR expression is ubiquitous41. Thus, AUF1 and TIAR might act to fine-tune MYC levels, with important biological consequences. For example, the level of MYC sets the balance between proliferation and differentiation. Tuning MYC expression to low levels allows differentiation, whereas high levels favor proliferation. Deregulation of this balance seems to be a major contributor to formation of many tumors18, 19.
How does AUF1 control MYC translation? Our data suggest two mechanisms (Fig. 7). (i) AUF1 may serve to limit binding of TIAR, a translational suppressor, to the MYC ARE, thus promoting MYC translation indirectly. For example, knockdown of AUF1 increased TIAR binding and subsequent translation suppression in an ARE-dependent fashion (Fig. 4). By contrast, AUF1 overexpression reduced association of TIAR with the MYC ARE and permitted elevated translation (Fig. 5). In addition, AUF1 might affect association of other ARE-binding proteins with MYC mRNA for translational control. However, future work will be required to adequately address this issue. (ii) AUF1 expression seems to directly promote translation as well. Knockdown of TIAR elevated MYC levels about two-fold (Fig. 4c,d). However, knockdown of TIAR together with AUF1 reversed the effect. We interpret this result to mean that AUF1 does not simply block binding of the TIAR suppressor protein to MYC mRNA to promote its translation. Then, how might AUF1 directly promote MYC translation? AUF1 immunoprecipitation experiments revealed that it forms a complex with translation initiation factor eIF4G, PABP, heat shock proteins hsc70-hsp70 and other unidentified proteins in cells15. Furthermore, a study of the hierarchy of protein-protein interactions between AUF1 and translation initiation complexes using in vitro binding assays28 has indicated that all four AUF1 isoforms can directly interact with the C-terminal region of eIF4G. They can interact with PAPB as well—both independently of, and in a complex with, eIF4G. These observations are consistent with our finding that overexpression of AUF1 isoforms, either singly or combined, stimulated MYC translation (Fig. 1). Future experiments will be required to fully elucidate the detailed molecular mechanisms by which AUF1 promotes MYC translation, however.
Figure 7: Model for AUF1 control of MYC translation.
See text for details. Mutually exclusive ARE binding by TIAR or AUF1 is shown for simplicity. However, regulated shifts in their competitive binding equilibrium for ARE co-occupancy could permit graded translation levels. Unidentified ARE-binding proteins not depicted probably contribute to MYC translation as well.
Full size image (38 KB)How does AUF1 promote translational control via the MYC ARE but degradation via other AREs? Do AREs contain embedded regulatory codes dictating translation versus degradation? If so, some AREs probably contain such codes for both types of control; after all, the MYC ARE does confer instability upon a reporter -globin mRNA34. However, other AREs may dictate translational control but not mRNA degradation. This might help to explain why some ARE-like sequences predicted to act as destabilizing elements do not. For example, (A+U)-rich regions in the KROX20 and ZIF268 mRNAs (which encode transcription factors) resemble class I AREs (that is, nonoverlapping AUUUA pentamers), yet they do not confer rapid degradation upon a reporter -globin mRNA34. It is tempting to speculate that the KROX20 and ZIF268 sequences in question might instead permit translational regulation. Clearly, much additional work will be required to resolve these questions.
In conclusion, we have identified AUF1 as a previously uncharacterized translational regulator of MYC mRNA. The balance of AUF1 versus TIAR expression could serve to fine-tune MYC translation to either a higher rate, in response to growth factor signals for proliferation, or a lower rate, in response to differentiation signals. Observations that phosphorylation of AUF1 controls its functions suggest that MYC expression may be linked to signaling pathways previously not considered, both in normal and in neoplastic cells42, 43, 44, 45.
Methods
Cell culture.
K562 (human chronic myeloid leukemia), THP-1 (human promonocytic leukemia), HeLa (human cervical carcinoma) and HT-29 (human colon carcinoma) cells (ATCC) were cultured at 37 °C in 5% (v/v) CO2 using RPMI-1640 or DMEM medium supplemented with 10% (v/v) FBS (GIBCO) and 2 mM glutamine (Invitrogen). Two K562 sublines, KmycB and KmycJ, which conditionally overexpress MYC, were maintained in the above RPMI-1640 medium with 100 
g ml-1 hygromycin B (Calbiochem). These sublines were generated by transfection of K562 cells with plasmid pHEBoMTmyc2.3 followed by clone selection46. Exogenous MYC expression in KmycB and KmycJ cells is under control of the mouse metallothionein I promoter and was induced by addition of 75 M ZnSO4 (Sigma) to the culture medium for 3 d.
RNA interference.
Plasmids encoding a control shRNA or shRNA directed against AUF1 will be described elsewhere. To exclude off-target effects of AUF1 shRNA, two siRNAs with AUF1 sequences differing from the shRNA were also used. The siRNAs correspond to coding region nucleotides 519–539 (5'-aag auu gac gcc agu aag aac-3') and 709–727 (5'-gau ccu auc aca ggg cga u-3'), respectively, of human AUF1 (GenBank NM_031370.2)7. The AUF1 shRNA and siRNAs described above target regions common to all four AUF1 isoforms. siRNA directed against TIAL1 (the gene that encodes TIAR, also called TIAR) targets coding region nucleotides 763–783 of human TIAR (5'-aag ggc uau uca uuu gtc aga-3'; GenBank M96954.1)27. MYC siRNA targets coding region nucleotides 607–627 of human MYC (5'-aac aga aau guc cug agc aau-3'; GenBank X00198.1)47. An siRNA containing nucleotides randomly arranged (5'-aac ugg gua agc ggg cgc aaa-3') was used as a negative control48. siRNA duplexes contain dTdT 3' overhangs and were chemically synthesized by Dharmacon Research. siRNA duplexes were dissolved in 1 
universal RNA oligo buffer (20 mM KCl, 6 mM HEPES-KOH (pH 7.5), 0.2 mM MgCl2). For RNAi, 5 106 cells were transfected with 10 g shRNA plasmid or 100 nM (final concentration) siRNA by electroporation using a Gene Pulser (Bio-Rad). Electroporation parameters were 280 mV, 1,050 F. Transfected cells were maintained in culture medium without antibiotics for 2–3 d to permit knockdown before assays. Knockdown efficiency was assessed by western blot analysis.
Western blot analysis.
Briefly, cytoplasmic or whole-cell lysates were prepared using the CelLytic NuCLEAR extraction kit (Sigma). Phase-contrast microscopy was used to monitor cell lysis. Protein concentration in lysates was quantified by Bradford assay using protein assay reagent from Bio-Rad. Lysate protein (40 g) was size-fractionated by SDS-PAGE and transferred onto nitrocellulose membranes (Fisher). The antibodies used and their dilutions were as follows: AUF1 (Upstate), 1:3,000; MYC (CalBiochem), 1:200; TIAR (Santa Cruz Biotechnology), 1:200; -tubulin (Sigma), 1:8,000. -tubulin served as the loading control. Blots were developed using the SuperSignal West Chemiluminescent Substrate kit (Pierce) according to the manufacturer's protocol. Signals were detected by exposure to X-ray film. Results were quantified using a DC120 Zoom Digital Image system (Kodak). Protein abundances in controls were arbitrarily set to 100%.
Luciferase reporter constructs.
pGL3-Promoter, a firefly luciferase expression plasmid, was purchased from Promega. To generate pGL3-Luc-MYC ARE, a 400-base-pair (bp) segment of the MYC 3' UTR 75 bp downstream of the stop codon was amplified by PCR from plasmid pMycSD3 (ref. 23) and subcloned into plasmid pGL3-Promoter at the XbaI site. The sequence of the MYC 3' UTR segment linked to the luciferase plasmid (pGL3-Promoter) corresponds to nucleotides 7290–7693 of the human MYC gene (GenBank X00364, J00120). This 400-bp region contains the ARE and spans both polyadenylation sites. Sense and antisense orientations of the MYC ARE in the luciferase reporter were identified by DNA sequencing.
Dual luciferase reporter assay.
Luciferase activity was examined using the Dual Luciferase Reporter Assay kit (Promega) following the manufacturer's instructions. Briefly, 5 106 cells were cotransfected with either pGL3-Luc-MYC ARE, pGL3-Luc-MYC ARE (antisense) or pGL3-Promoter; 10 g shRNA plasmid or 100 nM siRNA; and plasmid pRL-SV40 encoding Renilla luciferase (Promega) by electroporation, as described above. pRL-SV40 served as an internal control. A broad range of amounts of luciferase plasmids (10 ng–10 g) were tested in transfections, and comparable results were obtained across this range (data not shown). Luciferase activity in cytoplasmic extracts was measured 72 h after transfection using a TD-20/20 luminometer (Turner Designs). Firefly luciferase activity was normalized to Renilla luciferase activity in the same cell extract and expressed as a ratio of firefly/Renilla luciferase activity.
Quantitative real-time reverse-transcription PCR.
MYC, luciferase and GAPDH mRNA levels were examined by qRT-PCR using the QuantiTect Probe RT-PCR kit (Qiagen) with the MX3005P Multiplex Quantitative PCR System (Stratagene) as described49. Amplification plots, such as that shown in Figure 2c, represent fluorescence, Rn, versus cycle. Rn is the ratio of the fluorescence emission intensity of the reporter dye to the fluorescence emission intensity of the passive reference dye. Rn is the magnitude of the signal generated by the specified set of PCR conditions (Rn = Rn - baseline). qRT-PCR products were visualized by 2% (w/v) agarose gel electrophoresis to monitor quality. Where possible, GAPDH mRNA served as an internal control for normalization.
mRNP immunoprecipitation.
Immunoprecipitation of endogenous RNA–protein complexes was done by the protocol described49. Briefly, fresh cytoplasmic lysates were precleared with preimmune serum. Precleared lysate proteins (3 mg) were incubated for 3 h at 25 °C with protein A–Sepharose beads (Sigma) precoated with antibody. For AUF1 immunoprecipitation, 10 l antiserum specific for human AUF1 or preimmune serum was used10. For TIAR, 30 g goat polyclonal anti-human TIAR antibody (C-18; Santa Cruz Biotechnology) or normal goat IgG (BD Life Sciences) was used. A reaction containing preimmune serum or normal goat IgG served as a negative control. Half of the beads were washed and boiled for western blot analyses. The other half were used for mRNA extraction of the immunoprecipitate, followed by RNase-free DNase I (Roche) and proteinase K (Ambion) digestions, phenol-chloroform extraction and ethanol precipitation. qRT-PCR was used to assess the abundance of AUF1- or TIAR-associated MYC and luciferase–MYC ARE reporter mRNAs, as described above. The qRT-PCR products were visualized by 2% (w/v) agarose gel electrophoresis.
Polyribosome gradient analyses.
Polyribosome profile analyses were done as described50 with minor modifications. Briefly, 5 107 cells were cultured in medium with 0.1 mg ml-1 cycloheximide (Calbiochem) at 37 °C for 10 min, then harvested and washed twice with cold PBS containing 0.1mg ml-1 cycloheximide. Cell pellets were resuspended in an equal volume of PEB lysis buffer (20 mM Tris-HCl (pH 7.5), 100 mM KCl, 5 mM MgCl2, 0.3% (v/v) Triton X-100, 0.1 mg ml-1 cycloheximide, 1 mM DTT, 200 units ml-1 RNase Out (Invitrogen), 40 l ml-1 complete protease inhibitor mixture (Roche Applied Science)) on ice for 10 min. Lysed cells were centrifuged at 13,000g for 10 min at 4 °C. Protein concentration of cytoplasmic lysate was measured by Bradford assay. Approximately 2.5 mg cytoplasmic lysate proteins were layered on top of linear 10%–50% (w/v) sucrose gradients (11 ml). Tubes were centrifuged in a Beckman SW41 rotor at 260,500g for 90 min at 4 °C. Polyribosome profiles were obtained by absorbance measurements at 254 nm during fraction collection (1 ml each). RNA in 300 l of each fraction was extracted using the RNeasy kit (Qiagen) and treated with DNase I included in the kit. RNA samples were assayed for MYC and GAPDH mRNA levels by qRT-PCR as described above. For analysis of the relative distributions of MYC and GAPDH mRNAs in polyribosome gradients, Ct values from individual fractions 2–11 were each subtracted from the Ct value from fraction 1, as fraction 1 had the largest Ct values (that is, the lowest MYC and GAPDH mRNA abundances). The resulting Ct numbers were converted into fold differences using software included with the MX3005P Multiplex Quantitative PCR System. The abundance of each mRNA as a percentage of the total from all 11 fractions was then calculated. Because GAPDH mRNA is not a binding target of AUF1 (B.L. and G.B. unpublished data), its abundance was used as a control for estimation of relative distributions of MYC mRNA in polyribosome gradients after AUF1 knockdown or overexpression. To measure AUF1 and -tubulin protein abundances in fractions, we used 12 l of each fraction for western blot analyses as described above. -tubulin is a cytoplasmic protein not associated with ribosomes.
Cell proliferation assay.
Equal numbers of K562 cells were transfected with a plasmid expressing either the control shRNA or AUF1 shRNA and cultured in 96-well plates for 3 d to permit AUF1 knockdown. Cell proliferation was examined by methanethiosulfonate (MTS) assay using the CellTiter 96 AQueous One Solution Cell Proliferation Assay kit (Promega) as described previously48. Absorbance at 490 nm is directly proportional to the number of viable cells in culture. In some experiments, a K562 subline (KmycB) expressing MYC under control of the Zn2+-inducible metallothionein promoter was transfected with either control or AUF1 siRNA and maintained in culture medium with or without 75 M ZnSO4 for 3 d. ZnSO4 induces expression of MYC in the KmycB subline46. Cells were then used for western blot analyses and MTS assays.
Statistical analyses.
Data are depicted as mean s.d. from at least three independent experiments. Exact n values are provided in figure legends. One-way analysis of variance (ANOVA) and unpaired two-way t-test were done using GraphPad Instat 3.0. The Dunnett multiple comparisons post-test was done if a significant F ratio was obtained in ANOVA. P < 0.05 was considered statistically significant. Exact P-values are provided in each figure legend.
Note: Supplementary information is available on the Nature Structural & Molecular Biology website.
Author contributions
B.L. designed and performed experiments, interpreted experimental data and wrote the paper. Y.H. performed experiments. G.B. designed experiments, interpreted experimental data and wrote the paper. All authors approved the final version of the paper.