Transcriptional Regulation of APOBEC3G, a Cytidine Deaminase That Hypermutates Human Immunodeficiency Virus*

Apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like 3G (APOBEC3G) is an antiretroviral deoxycytidine deaminase that lethally hypermutates human immunodeficiency virus type 1 (HIV-1) but is itself neutralized by the HIV-1-encoded viral infectivity factor. Accordingly, APOBEC3G occurs specifically in human T lymphocytic cell lines that contain this antiviral defense, including H9. Since the substrate specificities of related cytidine deaminases are strongly influ-enced by their intracellular quantities, we analyzed the factors that control APOBEC3G expression. The levels of APOBEC3G mRNA and protein were unaffected by treatment of proliferating H9 cells with interferons or tumor necrosis factor- (cid:1) but were enhanced up to 20-fold by phorbol myristate acetate. This induction was mediated at the transcriptional level by a pathway that re-quired activation of the protein kinase C (cid:1) / (cid:2) I isozyme (PKC), mitogen-activated protein kinase kinase (MEK) 1 and 2, and extracellular signal-regulated kinase (ERK). Correspondingly, induction of APOBEC3G was blocked by multiple inhibitors that act at diverse steps of this pathway. The PKC (cid:1) / (cid:2) I/MEK/ERK pathway also controlled basal levels of APOBEC3G mRNA and protein, which consequently declined when cells were treated with these inhibitors or arrested in the G 0 state of the cell cycle by serum

Although APOBEC3G occurs in T lymphocytes and in several leukemic T cell lines including HUT78 and its derivative H9 that contain this antiviral defense system, it is absent in many other cell lines that have been examined (7,19). Consequently, these other cell lines are permissive for replication of HIV-1(⌬vif) that contains an inactivating mutation or deletion of the vif gene (20 -22). Expression of APOBEC3G in these permissive cell lines converts them to the nonpermissive phenotype, as determined by their ability to efficiently inactivate HIV-1(⌬vif) but not wild-type HIV-1 (4,(7)(8)(9)(10)(11)(12)(13)(14)19). Recently, another member of the cytidine deaminase family, APOBEC3F, was also shown to have antiretroviral activity (23,24). Similarly, the RNA-specific adenosine deaminases, which are induced by interferons in response to double-stranded RNA, inactivate several RNA viruses (25).
The factors that control expression of APOBEC3G have not previously been investigated. Furthermore, it is unknown whether the APOBEC3G that occurs naturally in T lymphocytes is constitutively active or whether its antiviral function is induced by infection or by signal transduction processes such as the classical innate antiviral responses involving type I interferons (INF) and tumor necrosis factor-␣ (TNF-␣) (25,26). APOBEC3G occurs in a family of cytidine deaminases that includes APOBEC1, -2, -3A, -3B, -3C, -3E, -3F, and activationinduced cytidine deaminase. It is encoded on chromosome 22 in a region that also contains the genes or pseudogenes for APOBEC3A to 3F (27,28). Previously, it was shown that phorbol esters cause increases in the amounts of APOBEC3A and APOBEC3B mRNAs by an unknown mechanism. Consequently, these enzymes were initially termed phorbolin-1 and phorbolin-2, respectively (29,30). Phorbol myristate acetate (PMA) activates several protein kinase C (PKC)-dependent and -independent intracellular signaling pathways (31)(32)(33). The most extensively studied PKC-dependent pathway involves activation of Ras and the downstream kinases Raf, mitogenactivated protein kinase kinase 1 and 2 (collectively termed MEK), and the extracellular signal-regulated kinase (ERK) (34 -36). Activated ERK stimulates specific transcription factors that have been shown to up-regulate expression of a diverse array of genes in different tissues (35,37). ERK is a member of the mitogen-activated protein kinase (MAPK) fam-ily, which also includes JNK and p38. The latter are not activated by PMA but respond to stress stimuli and inflammatory cytokines (37,38).
Specific factors have been shown to regulate the expression and to control the functions of APOBEC1 and activation-induced cytidine deaminase (39 -43). The substrate specificity of the well studied APOBEC1 enzyme is controlled by APOBEC1 complementation factors, which recruit it to a specific site in the apolipoprotein B mRNA (41,43). It is only when APOBEC1 is expressed in excess of the APOBEC1 complementation factors that it deaminates other mRNAs (44,45). Interestingly, overexpression of APOBEC3G and other cytidine deaminases can cause mutagenesis and cellular toxicity (45)(46)(47). Thus, there is precedence for the hypothesis that cellular factors and expression levels of cytidine deaminases can regulate their functions.
Here we describe our initial characterization of the factors that control APOBEC3G expression levels in T cells. Interestingly, APOBEC3G mRNA synthesis is dynamically controlled by a protein kinase cascade that is induced by extracellular signals and that is inhibited when cells become arrested in the G 0 stage of the cell cycle. Northern Blot Analysis and qRT-PCR-RNA extraction and Northern blot analysis were described previously (48). [ 32 P]-labeled cDNA probes were used to detect APOBEC3G, protein kinase R (PKR), or the S2 ribosomal protein mRNAs (used as a loading control). Real-time quantitative reverse transcription PCR (qRT-PCR) was performed according to standard protocols (49). Briefly, 1 g of total RNA was reverse-transcribed using random hexamer primers and Superscript II reverse transcriptase (Invitrogen). The cDNAs of glyceraldehyde-3phosphate dehydrogenase and APOBEC3G were amplified using 2ϫ SYBR Green master mix (2ϫ PCR buffer, 4 mM MgCl 2 , 0.4 mM dNTPs, 0.005% SYBR Green, 2ϫ Rox reference dye, 16% Me 2 SO, 0.04 units/l Platinum Taq polymerase) and an ABI Prism 7700 sequence detection system (Applied Biosystems). The sensitivity of the PCR was tested by amplification of the target from serially diluted cDNAs generated from reverse transcription of human reference RNA (Stratagene). qRT-PCR amplification within samples was normalized using glyceraldehyde-3phosphate dehydrogenase amplification levels as an endogenous control. Each sample was assayed in triplicate using the primer pair 5Ј-T-CAGAGGACGGCATGAGACTTA-3Ј, 5Ј-AGCAGGACCCAGGTGTCAT-T-3Ј specific for APOBEC3G and the primer pair 5Ј-GAAGGTGAAGG-TCGGAGT-3Ј, 5Ј-GAAGATGGTGATGGGATTTC-3Ј for glyceraldehyde-3-phosphate dehydrogenase. Data analysis and calculations were done following the 2 -⌬⌬CT comparative method outlined in Ref. 62.
Western Blot Analysis-Western blot analysis was described previously (50). Briefly, cells were lysed with radioimmune precipitation buffer containing phosphatase inhibitors (50 mM Tris-Cl (pH 7.4), 1% Nonidet P-40, 0.1% sodium deoxycholate, 150 mM NaCl, 2 mM NaVO 3 , and 25 mM NaF) and Complete protease inhibitors (Roche Applied Science). Cell lysates were adjusted to equal protein concentrations by using Bradford reagent (Bio-Rad), and equal amounts were used for Western immunoblotting. APOBEC3G was detected using a rabbit peptide antibody (a gift from Warner Greene, University of California, San Francisco). MAPKs were detected using antibodies for MEK-1 and ERK-1 and phospho-specific antibodies for detecting the phosphorylated forms of both MEK and ERK (Santa Cruz Biotechnology). Activation of p38 was monitored using phospho-specific antibodies for p38 MAPK (a gift from Bruce Magun, Oregon Health and Science University). Equivalent loading of proteins was shown using a mouse antibody specific for ␣-tubulin (Sigma).

Control of APOBEC3G Expression by Specific Protein Kinase C Isozymes-Northern blot analyses in stringent conditions
and real-time quantitative reverse transcription-PCR (qRT-PCR) were both used to evaluate APOBEC3G mRNA levels in cells. Furthermore, because sequence similarities exist among APOBEC family members (27), we used oligonucleotide primers that were specific for APOBEC3G, and we cloned and sequenced multiple PCR products from independent reactions and confirmed that they all corresponded to APOBEC3G. As shown in Fig. 1, A and B, APOBEC3G mRNA levels in H9 cells were unaffected by treatments with INF-␣, -␤, or-␥, which were active as indicated by their inductive effects on PKR mRNA levels (Fig. 1B). Similarly, TNF-␣ had no effect on the APOBEC3G mRNA level, although it was also active in these cells as indicated by its ability to phosphorylate and activate p38 MAPK (data not shown). In contrast, PMA induced strong increases in APOBEC3G mRNA levels within 2 h that continued until at least 10 h after treatment. This effect was specific for PMA because an inactive phorbol ester, 4␣PMA, had no effect on APOBEC3G mRNA levels (Fig. 2B). Prolonged exposure to PMA causes elimination of PKC (33), which may explain the diminished effect of PMA on APOBEC3G expression levels at 24 h after treatment (Fig. 1).
PMA was previously shown to increase mRNA levels of APOBEC3A and APOBEC3B (29). PMA mimics the second messenger diacylglycerol that binds to and activates PKC. There are 11 known PKC isozymes, 9 of which are PMAsensitive (␣, ␤I, ␤II, , ␥, ␦, ⑀, , ) (33). To determine whether enhancement of APOBEC3G mRNA levels by PMA treatment is mediated by PKC, general PKC inhibitors (Calphostin C, PKC inhibitor 20-28) as well as inhibitors specific for PKC␣/␤I (Gö6976, Ro-32-0432) were incubated with H9 cells 1 h prior to and for the duration of PMA or a mock treatment. All PKC inhibitors abolished the induction of APOBEC3G mRNA by PMA ( Fig. 2A). Inhibitors that are highly selective for PKC␣/␤I isozymes were as effective as general PKC inhibitors at preventing PMA enhancement of APOBEC3G mRNA, suggesting that PKC␣/␤I isozymes are solely responsible for this effect.
MEK Regulates APOBEC3G mRNA Levels-Activation of PKC␣/␤I is critical for the regulation of mRNA levels for many genes in numerous cell lines (34,35,37). The major downstream target of PKC␣/␤I is MEK (36,38). To evaluate the possible role of MEK as a downstream target of PMA treatment, H9 cells were treated for 1 h with specific MEK inhibitors U0126 or PD98059 (51,52) prior to and during PMA treatment. These MEK inhibitors efficiently blocked the PMAinduced increase in APOBEC3G mRNA (Fig. 2, B and C). Indeed, APOBEC3G mRNA levels in H9 cells treated with U0126 or PD98059 declined below the basal level within 4 h after treatment and continued to decline over a 24-h period (Fig. 2, B and C), suggesting that MEK-dependent signal transduction is necessary both to maintain the basal level of APOBEC3G and to enhance the APOBEC3G mRNA level in response to PMA.
We also did Western immunoblot analyses of H9 cell extracts using a previously characterized antiserum made to a peptide corresponding to the carboxyl-terminal 16 amino acids of APOBEC3G. This antiserum is highly specific for APOBEC3G (results not shown) (7) and detected a single protein in the cell extracts in relative amounts that correlated closely with the cellular APOBEC3G mRNA levels (Fig. 3). Serum starvation arrests cultured cells at the G 0 /G 1 transition of the cell cycle and causes dephosphorylation and consequent inactivation of MEK, whereas progression to G 1 coincides with MEK activation (7,(53)(54)(55). Accordingly, serum starvation of H9 cells caused a reproducible decline in the APOBEC3G protein content, and PMA treatment of serum-starved H9 cells caused an induction of APOBEC3G protein that was blocked by the MEK inhibitors PD98059 and U0126 (Fig. 3, left panel). Although these treatments had no effect on the total cellular quantity of MEK, PMA induced the phosphorylation of MEK, and this was also inhibited by PD98059 and U0126 (Fig. 3, right panel). This result is consistent with previous evidence that these compounds can inhibit the phosphorylation of MEK as well as the activity of the phosphorylated MEK (51). The only known downstream substrates of phosphorylated MEK are ERK1/2 (56). As shown in Fig. 3 (right panel), serum starvation caused almost complete elimination of phosphorylated ERK, whereas PMA induced its phosphorylation in these serum-starved cells. This induction was prevented by the MEK inhibitors PD98059 and U0126 (Fig. 3, right panel). In contrast, total cellular ERK levels were unaffected by these treatments.
PMA Induces Increased APOBEC3G mRNA Synthesis-Activation of ERK has been reported to increase levels of mRNAs either by transcriptional enhancements or by stabilization of mRNAs (34,37). To distinguish between these possibilities, H9 cells were mock-treated or treated for 8 h with PMA before actinomycin D was added to block further transcription. Aliquots of cells were taken at subsequent times to monitor APOBEC3G mRNA stability. At the time of actinomycin D addition, the level of APOBEC3G mRNA was approximately four times higher in the PMA-treated cells than in the untreated control cells. Consequently, the blot exposures were normalized to show equal intensities of the APOBEC3G mRNAs at the time of actinomycin D addition. As shown in Fig.  4A, no significant differences in APOBEC3G mRNA stability were observed over 16 h in the presence or absence of PMA. Therefore, we conclude that PMA does not function by stabilizing APOBEC3G mRNA. To evaluate the synthesis of APOBEC3G mRNA in the presence of PMA, H9 cells were cultured in the absence of serum for 24 h followed by treatments with the transcription inhibitor actinomycin D or the protein synthesis inhibitor cyclohexamide, in combination with specific MEK inhibitors. After 1 h, cells were stimulated with PMA. PMA caused an increase in APOBEC3G mRNA that was blocked by all of the inhibitors (Fig. 4B). These results suggest that PMA induces an increase in APOBEC3G mRNA synthesis by a pathway that requires continued protein synthesis and activated MEK.
Similar inductive effects of PMA on APOBEC3G levels were seen in several other proliferating T cell lines that contained basal amounts of APOBEC3G in the absence of PMA. However, PMA did not induce significant amounts of APOBEC3G expression in cells that had very low or negligible basal amounts of this protein, implying that PMA alone may be unable to activate fully repressed APOBEC3G genes (results not shown). In agreement with this observation, we also found in a preliminary study that PMA did not convert permissive cells to the nonpermissive phenotype. Conversely, we found that H9 cells did not become permissive when they were starved for serum or treated with MEK inhibitors for 24 h in conditions that cause reductions in intracellular levels of APOBEC3G. We consider the latter studies preliminary because these treatments caused only modest reductions in the intracellular amounts of APOBEC3G (Fig. 3), presumably because APOBEC3G mRNA and protein turn over relatively slowly (Figs. 2 and 4) (7, 10). DISCUSSION These studies provide initial evidence concerning the dynamic regulation of APOBEC3G mRNA and protein levels in the human T lymphocytic cell line H9. Unlike many other antiviral factors (e.g. adenosine deaminases, ribonuclease L, Mx GTPase, PKR) (25), the syntheses of APOBEC3G mRNA and protein were unaffected by type 1 or type 2 interferons (Fig. 1). These interferons were active in H9 cells as indicated by their abilities to induce the synthesis of PKR. In addition, APOBEC3G was unaffected by TNF-␣, which was also active in these cells (Fig. 1). In addition, APOBEC3G mRNA levels do not change in response to HIV-1 infection (7,9). In striking contrast, the quantity of APOBEC3G mRNA was enhanced ϳ20-fold by treatment of the cells with PMA ( Figs. 1 and 2). This induction was mediated by the PKC␣/␤I isozymes and by the downstream protein kinase cascade involving MEK and ERK, and it was consequently inhibited by the PKC␣/␤I inhibitors Gö6976 and Ro-32-0432 and by the MEK inhibitors U0126 and PD98059 (Figs. 2-4). These inductive and inhibitory effects were seen using Northern blot analyses in stringent hybridization conditions as well as qRT-PCR assays using primers that were specific for APOBEC3G (Figs. 1, 2C, and 4) and were confirmed by protein immunoblot analyses using an antipeptide antiserum made to the specific carboxyl-terminal region of APOBEC3G (Fig. 3) (7).
Interestingly, the basal expression of APOBEC3G also appeared to depend on this same protein kinase cascade. Thus, the basal level of APOBEC3G mRNA was reduced in the pres- ence of these inhibitors (Fig. 2, B and C). In accordance with these conclusions, it is known that MEK becomes dephosphorylated when cells enter the G 0 state of the cell cycle (53)(54)(55), and we verified this using serum-starved cells as shown in Fig.  3. This treatment caused accumulation of the cells in the G 0 state and resulted in dephosphorylation of the downstream target ERK (Fig. 3). PMA treatment of these growth-arrested cells induced the phosphorylation of MEK and of ERK and the accumulation of APOBEC3G mRNA and protein in a manner that was inhibited by the MEK inhibitors U0126 and PD98059 (Figs. 3 and 4B). Considered together, these data strongly indicate that the expression levels of APOBEC3G mRNA and protein are regulated by a protein kinase cascade involving PKC␣/␤I, MEK1/2, and ERK1/2 and by the interactions of this cascade with the normal cell cycle. These results may potentially explain the increase in APOBEC3G previously reported to occur when resting T lymphocytes are activated with interleukin-2 and phytohemagglutinin (7).
Our data further suggest that the effects of PKC␣/␤I and MEK are mediated by changes in synthesis of APOBEC3G mRNA rather than by alteration of the mRNA stability (Fig. 4,  A and B). Thus, the rate of APOBEC3G mRNA degradation was not significantly altered despite the substantial increase in APOBEC3G mRNA caused by the PMA-dependent activation of PKC␣/␤I. Previous evidence has shown that the MAPK pathway implicated by our results can activate numerous transcription factors including cAMP-response element-binding protein (CREB), NFB, AP-1 (c-Jun, c-Fos), c-Myc, Ets-1, and Elk-1 (34). A domain search of the 5Ј-untranslated region of APOBEC3G (accession number NM_021822) using web-based search engines, TFSEARCH (57) and TESS (58), revealed several binding sites for Ets-1, c-Myc, and Elk-1. Further investigations will be needed to determine whether these putative binding sites are bound by transcription factors and whether any of these ERK-regulated transcription factors are responsible for the PMA-induced increase in APOBEC3G transcription implied by our data.
It was previously shown that APOBEC3A and APOBEC3B are induced by phorbol esters (29). Our results extend this to APOBEC3G, which is encoded in the same gene cluster on chromosome 22 (27). Although we presume that the mechanisms of the APOBEC3A and APOBEC3B inductions may be similar to that described here for APOBEC3G, the PMA induction pathways for these other cytidine deaminases have not been investigated. In any case, the close linkage of these genes that are activated by PMA raises the possibility that they might be coordinately regulated by a PKC␣/␤I/MEK/ERK-dependent locus control region as well as by promotor-specific transcription factors that influence their tissue-and cell-specific patterns of expression. This differential pattern of expression is clearly important in the case of APOBEC3G, which is present only in some lymphoid and myeloid cell lines (27).
It is unknown whether APOBEC3G has a normal cellular function in addition to its role in inhibiting the retroviruses and retroelements that continually invade the genomes of mammals. However, in view of its powerful antiretroviral activity, it is intriguing that APOBEC3G is so strongly and dynamically regulated by a protein kinase cascade that is known to be activated by extracellular signaling molecules (34,35). The functions and substrate specificities of other cytidine deaminases can be controlled by their expression levels and interactions with accessory factors (42, 43, 45, 59 -61). Further studies will be needed to learn how these or other regulatory pathways might control the antiviral functions of APOBEC3G.