APRIL Binding to BCMA Activates a JNK2–FOXO3–GADD45 Pathway and Induces a G₂/M Cell Growth Arrest in Liver Cells

APRIL (TNFSF13) (1) is a member of the TNF superfamily, produced by macrophages, monocytes, and dendritic cells (2). APRIL and BAFF (TNFSF13B) (3) bind to two receptors (4–7), BCMA (TNRSF17) (8, 9) and TACI (TNFRSF13B) (10), but with a different affinity (11–13); BAFF also binds to its specific receptor, BAFFR (TNFRSF13C) (14, 15). APRIL has been reported to bind also the proteoglycan syndecan-1 (CD138) (16, 17), and a not-yet-identified APRIL receptor has been advanced, as well (18, 19). This complex system of two ligands and (at least) three receptors was considered exclusive to B lymphocytes; however, recently, all or some of these five molecules have been identified in mesenchymal cells (20), normal tissues (20–24), and epithelial tumors (20, 21, 24–26), suggesting additional non–immune cell-related functions for these ligands and receptors, in both physiological and malignant tissues. APRIL has also been found to be expressed in some normal germinal-center B cells, myeloid and dendritic cells, malignant B cells and cell lines, activated T cells, glioblastoma cell lines, and the placenta (1, 18, 27–33). BAFF has also been detected in some of these tissues, but no significant differences between normal and neoplastic tissue are usually reported; the expression of its cognate receptor BAFFR is either missing or occurs independently of tumor evolution (20, 21, 24–26). Binding of APRIL or BAFF to BCMA or TACI triggers diverse signaling pathways in engineered cells, including nuclear translocation of NF-κB, activation of p38 mitogen-activated kinase and JNK phosphorylation for BCMA (34), and NF-κB nuclear translocation and JNK phosphorylation for TACI (11).

Previous investigations identified the presence of BCMA transcripts in liver tissue (35). In this article, we report that APRIL and BCMA are expressed in hepatocellular carcinoma (HCC) specimens and two HCC cell lines (HepG2, Hep3B), as well as in normal liver and in a small series of human HCCs, whereas TACI is absent. We further show that APRIL, which is usually an inducer of cell growth, binds to BCMA and inhibits cell proliferation in HCC cells, through G₂/M cell cycle arrest, whereas BAFF has no effect on the proliferation of these cells. These data suggest that HepG2 and Hep3B cells could be a convenient model for analyzing specific events mediated exclusively by BCMA in nonengineered cells. We further report that the APRIL/BCMA antiproliferative action is mediated via a novel signaling pathway that involves JNK2 phosphorylation, FOXO3A activation, and GADD45 transcription.

Human HCC-derived lines HepG2 and Hep3B were purchased from DSMZ (Braunschweig, Germany); they were cultured in RPMI 1640 and DMEM, respectively, supplemented with 10% FBS (Invitrogen, Carlsbad, CA), at 37°C, 5% CO₂. Wortmannin, LY-294002 (PI3K/Akt inhibitors), and SP600125 (pan-JNK inhibitor) were purchased from Calbiochem (San Diego, CA). All other chemicals were purchased from Sigma-Aldrich (St. Louis, MO), unless stated otherwise. To exclude the possibility of a batch/production effect on our experiments, we have used for almost all experiments recombinant APRIL purchased from either Abcam (Cambridge, U.K.) or ROCHE Hellas (Marousi, Greece).

Twelve histologically verified HCC samples were retrieved from the Pathology Department (University of Crete, School of Medicine) tissue archives, after formal approval from the University Hospital Research and Ethics Committee. Seven normal liver samples acquired from autopsies after car accidents, with consent of the families, were also used. Immunohistochemistry in pathology samples and in HCC cell lines was performed as previously described (24, 36), in 3-μm sections. Abs against APRIL (ALX-804-149-C100, mouse monoclonal), BAFF (ALX-804-131-C100, rat monoclonal), and BCMA (ALX-804-151-C100, rat monoclonal) were from Alexis (Lausanne, Switzerland). The Ab against BAFFR (AF1162, goat polyclonal) was from R&D Systems (Minneapolis, MN), whereas the Ab against TACI (Sc-80335, mouse monoclonal) was from Santa Cruz Biotechnology (Santa Cruz, CA). Detection systems were TL-125-AL, Lab Vision Fast Red from Thermo Fisher Scientific (Fremont, CA), or K1500 (DAB) and K0689 (Fast Red) from Dako (Glostrup, Denmark). Positive controls according to our previous studies in human cells and tissues [described in previous publications (20, 24, 36, 37)] are presented in Supplemental Fig. 1A.

A hybrid construct encoding the signal peptide sequence of human H chain γ 1 fused to the extracellular part of human BCMA, connected in-phase with the Fc segment of human IgG1, was introduced into the plasmid p7055, as described previously (18). The secreted BCMA-Fc protein was collected in the supernatant and purified on protein A–agarose beads.

RT-PCR was performed as described previously (24). Positive controls were run in parallel with samples included in the study. We used adipose-derived mesenchymal stem cells as positive controls of BAFF and APRIL (20) and isolated human lymphocytes as positive controls for BAFFR, TACI, and BCMA. Quantitative RT-PCR (qRT-PCR) was performed as described previously (38). Changes were normalized according to 18S RNA expression. The primers used for the study were as follows: BAFF forward: 5′-TTC TAG GGC ACT TCC CCT TT-3′, reverse: 5′-CTC AAG ACT GCT TGC AAC TGA-3′; APRIL forward: 5′-TCT CCT TTT CCG GGA TCT CT-3′, reverse: 5′-CCA GAA TGG GGA AGG GTA TC-3′; BAFFR forward: 5′-AGG ACG CCC CAG AGC C-3′, reverse: 5′-AGT GTC TGT GCT TCT GCA GG-3′; TACI forward: 5′-AGT GAA CCT TCC ACC AGA GC-3′, reverse: 5′-CTC TTC TTG AGG AAG CAG GC-3′; BCMA forward: 5′-GTC AGC GTT ATT GTA ATG CAA GTG T-3′, reverse: 5′-TCT TTT CCA GGT CAA TGT TAG CC-3′; MCM2 forward: 5′-CAC CAC GTA CCT TGT GCT TG-3′, reverse: 5′-AAG GAT CAG CAG ATC GGA GA-3′; MCM4 forward: 5′-ACG TTT TGC ATC CGT TTT C-3′, reverse: 5′-CCT GGG GAC AGA GTG AAT GT-3′; MCM5 forward: 5′-AGG GAT CTT CAC CAG GTG TG-3′, reverse: 5′-GCC AAG GAG GTA GCT GAT GA-3′; MCM6 forward: 5′-CGA GGG AAT AGA CAC GAT CA-3′, reverse: 5′-CAT GTC CCG ATT CGA TCT CT-3′; PCNA forward: 5′-TCA GGT ACC TCA GTG CAA AAG-3′, reverse: 5′-TGC AAG TGG AGA ACT TGG AA-3′; RPA3 forward: 5′-CTC CCT TAG CTG CGT CAC TC-3′, reverse: 5′-GAG AAA GTG CTC AAG AAT GTT CA-3′; POLE2 forward: 5′-TTT TGC AGA AGT CTT CAC AGA TG-3′, reverse: 5′-GCA GAA GGT TGG TTT GAA GA-3′; GADD45 forward: 5′-GCA GGA TCC TTC CAT TGA GA-3′, reverse: 5′-CTC TTG GAG ACC GAC GCT G-3′; JNK1 forward: 5′-GCC AGA CCG AAG TCA AGA AT-3′, reverse: 5′-CAA GCA CCT TCA TTC TGC TG-3′; JNK2 forward: 5′-TTC AGG GTG CAG TCT GAT TTC-3′, reverse: 5′-TTC TTT ACC AGA TGC TTT GTG G-3′; 18S RNA forward: 5′-ATG GTC AAC CCC ACC GTG T-3′, reverse: 5′-TTC TGC TGT CTT TGG AAC TTT GTC-3′. They were selected from qPrimer Depot (qPrimerDepot, http://primerdepot.nci.nih.gov) and synthesized by VBC Biotech (Vienna, Austria).

Centrifugation-cleared supernatants from cells incubated for 24 h in six-well plates were collected, and cells were lysed and used for protein determination-normalization. APRIL and BAFF were assayed on pooled, 5×-concentrated media, using ELISA kits (Bender MedSystems, Vienna, Austria), according to the manufacturer’s instructions. The effect of the medium was systematically subtracted. Results are expressed as nanograms per milligram of total cellular proteins (measured with the Bradford Assay) per 24 h.

Short hairpin RNA (shRNA) against JNK1 and JNK2 was prepared with the use of the psiRNA-h7SKGFPzeo Kit (Invivogen, San Diego, CA), according to the manufacturer’s instructions. Briefly, psiRNA-h7SKGFPzeo plasmid was digested with BbsI (New England Biolabs, Ipswitch, MA) and was gel-purified with Extract II columns (Macherey-Nagel, Duren, Germany). The following oligonucleotides were used: JNK1 oligo1: 5′-ACC TCG AGT CGG TTA GTC ATT GAT AGT CAA GAG CTA TCA ATG ACT AAC CGA CTC TT-3′; JNK1 oligo2: 5′-CAA AAA GAG TCG GTT AGT CAT TGA TAG CTC TTG ACT ATC AAT GAC TAA CCG ACT CG-3′; JNK2 oligo1: 5′-ACC TCG AGC AGT TAG AGT AGG TGA ATT CAA GAG ATT CAC CTA CTC TAA CTG CTC TT-3′; JNK2 oligo2: 5′-CAA AAA GAG CAG TTA GAG TAG GTG AAT CTC TTG AAT TCA CCT ACT CTA ACT GCT CG-3′; scrambled shRNA oligo-1: 5′-ACC TCG GGT ATT TAG GCT ACG ATA GTT CAA GAG ACT ATC GTA GCC TAA ATA CCC TT-3′; scrambled shRNA oligo-2: 5′-CAA AAA GGG TAT TTA GGC TAC GAT AGT CTC TTG AAC TAT CGT AGC CTA AAT ACC CG-3′. Annealed oligonucleotides (95°C for 5 min and left to cool slowly to 35°C) were ligated with T4-DNA ligase (TAKARA, Otsu, Shiga, Japan) with the digested psiRNA-h7SKGFPzeo plasmid and used for the transformation of LyoComp GT116 cells, plated on Fast-Media Zeo X–Gal plates. After 24 h incubation at 37°C, white colonies were picked, and minipreps, positive for the insert, were prepared by incubation of isolated DNA, and digested with SpeI. A single positive miniprep was selected for expansion, and the presence of the proper insert was verified with sequencing.

Cells were grown for 6 d, with a change of medium and fresh APRIL or BAFF at day 3. Cell proliferation/viability was measured by the MTT assay. All substances were added to cultures 1 d after seeding (designated day 0). APRIL concentrations ranging from 1 to 200 ng/ml were used for these experiments.

For lentiviral shRNA studies, cells were plated at an initial density of 1.5 × 10⁴ cells in 24-well plates, with 0.5 ml medium per well. They were treated with MISSION anti-BCMA shRNA Lentiviral Transduction Particles (Sigma-Aldrich), according to the manufacturer’s instruction, in a hexadimethrine bromide environment (final concentration 8 μg/ml). The shRNA nontarget control particles were used as a negative control, and Mission TurboGFP Control Particles were used for assessment of transfection efficiency (>80% in all cases). The effectiveness of BCMA knockdown was verified with qRT-PCR (Supplemental Fig. 1B).

For JNK1 and two knockdown studies, cells were plated in either six-well or 24-well plates (2 × 10⁶ and 1.5 × 10⁴ cells/well, respectively) and left to adhere for 24 h. The medium was changed, and transfection was performed with a standard Lipofectamine 2000 protocol (Invitrogen; 0.8 μg DNA, 1 μl Lipofectamine 2000 in Optimem medium, for each well of a 24-well plate, scaling up for 6-well plates) and used, as described in 14Results, after 48 h. Transfection efficiency was >85%, as estimated based on GFP-positive cells. JNK1 and JNK2 knockdown was verified with qRT-PCR (Supplemental Fig. 1C) and Western blot (see 14Results).

Apoptosis of treated cells (APRIL, 200 ng/ml) was evaluated with (1) the Annexin VFITC Apoptosis Detection Kit and (2) the Caspase-3 Activity Assay (both from BD Biosciences, San Diego, CA), by flow cytometry on a FACSCalibur (BD Biosciences), according to the manufacturer’s instructions. Treated cells were compared with control (nontreated) samples.

For the assay of APRIL-induced modification of the cell cycle, treated (APRIL, 200 ng/ml) and control cells (1 × 10⁶ per well) were incubated for a minimum of 2 h at 4°C in the dark, with 1 ml of appropriate buffer (3.4 mM citrate, 25 mM NaCl, pH 7.6, 10% Nonidet P-40, 10 mg/ml RNase A, 10 mg/ml propidium iodide); measured by flow cytometry; and analyzed with the CELLQuest (BD) and ModFit LT (Verify Software, Topsham, MN) software. Experiments were performed three times in triplicates.

Cells, in 24-well plates, were transfected with 0.5 μg per well of pNFκB-Luc plasmid (Clontech, Mountain View, CA), carrying NF-κB response elements, in front of the 5′ end of the firefly luciferase gene, together with 0.5 μg per well of a Renilla luciferase vector (pRL-CMV; Promega, Fitchburg, WI), using Lipofectamine 2000 (Invitrogen, 1 μl per well) in Optimem medium. Cells were incubated for 24 h and then treated with APRIL (200 ng/ml, 24 h). Luciferase activity was assayed with a Dual-Luciferase Reporter 1000 Assay System (Promega), in a Berthold FB12 Luminometer (Bad Wildbad, Germany).

FOXO3A and E2F activation was performed with the Cignal Reporter Assay plasmids (SA Biosciences, Frederick, MD), according to the manufacturer’s instructions. Briefly, 10⁵ cells in 24-well plates were transfected with the reporter, or the negative or positive control (250ng/well), according to a standard Lipofectamine 2000 protocol. After a 24-h incubation, cells were treated (APRIL, 200 ng/ml), and 24 h later, luciferase activity was assayed.

Treated cells (APRIL, 200 ng/ml) in six-well plates were lysed in Laemmli's buffer; electrophoresis and blotting were performed according to standard procedures. The following Abs were used: phospho-JNK1/2 and total JNK (Abcam; 1/200), phospho-p38 [Thr¹⁸⁰, Tyr¹⁸²; (Cell Signaling Technology, Danvers, MA; 1/1000)], phospho-Akt and total Akt (Cell Signaling Technology; 1/1000), and anti-actin (Santa Cruz Biotechnology; 1/2000). At least three independent experiments were performed in each case.

HepG2 cells were treated with APRIL (200 ng/ml) for 2, 6, and 12 h. Total RNA was labeled and hybridized according to the Affymetrix protocol, using the Human Genome U133A Plus 2.0 Affymetrix GeneChip (54,675 transcripts). Normalization of signal and analysis were performed with Genespring GX V11.0 (Agilent, Foster City, CA). Transcripts modified by a factor of 2 in either direction (APRIL/vehicle) were extracted and further analyzed by use of the online resource gene set enrichment analysis (http://www.broadinstitute.org/gsea), followed by leading edge analysis (39). Finally, genes modified by APRIL after a 12-h incubation were analyzed for the reverse detection of putative transcription factors, by using the Web resource TFactS (http://www.tfacts.org) (40). Gene array data have been stored at the National Institutes of Health Gene Expression Omnibus repository (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE29375.

Statistical analysis was performed with the PASW v. 19.0 (SPSS, Chicago, IL) program. A statistical threshold of p < 0.05 was retained for significance.

APRIL and BCMA show an intense immunocytochemical expression in HepG2 and Hep3B cells (Fig. 1A). BCMA presents an intracellular and peripheral expression, compatible with a membrane or submembrane localization; in contrast, TACI immunostaining is absent, whereas BAFF and BAFFR are also expressed. A dotted expression of APRIL is observed, compatible with its mode of synthesis and secretion (2).

Immunocytochemical data were further confirmed by RT-PCR (Fig. 1B), revealing expression of APRIL, BCMA, BAFF, and BAFFR, whereas TACI expression was not evidenced. Relative quantification of gene expression by qRT-PCR analysis (41) revealed that BAFFR is expressed 10.98 ± 2.34-fold and 6.34 ± 1.23-fold more than BCMA in HepG2 and Hep3B cells, respectively. In addition, flow cytometry in nonpermeabilized cells (Fig. 1C) confirmed the absence of TACI and identified relatively low membrane expression of BAFFR and BCMA. Further, APRIL and BAFF were also found to be secreted to culture medium (HepG2, 2.35 and 5.08 ng/mg protein per 24 h, and Hep3B, 7.23 and 1.49 ng/mg protein per 24 h, respectively). This low level of secreted BAFF makes improbable its interaction with BCMA, in view of BAFF’s substantially lower affinity for the receptor (11–13).

APRIL and BAFF are considered cell proliferation promoters (42–45). However, in HCC cells, a differential effect was found: BAFF has no effect on cell proliferation (Supplemental Fig. 1D), whereas addition of APRIL results in a dose- and time-dependent, moderate inhibition of cell growth (Fig. 2A). This effect is specific, as the addition of Fc-BCMA, a soluble form of BCMA, results in a reversion of APRIL action (Fig. 2B). The exclusive implication of BCMA in this action is verified by the use of anti-BCMA shRNA. BCMA-specific, but not nontarget shRNA, made cells insensitive to APRIL (Fig. 2C). The APRIL effect is not due to apoptosis, assayed by Annexin V/propidium iodide staining (control and APRIL-treated cells, 2.61 and 1.38% in HepG2, and 2.96 and 0.75% in Hep3B cells, respectively; Supplemental Fig. 1E) or cleaved caspase-3 activity (4.82% versus 3.89% in HepG2 and 3.71% versus 3.52% in Hep3B cells for control and APRIL-incubated cells, respectively). We further report that APRIL induced a marginal but significant arrest of cells at the G₂/M phase of the cell cycle (Fig. 2D), a result that could explain the accumulated effect seen in our 3- and 6-d proliferation experiments. However, minor changes in the ratio phospho/total cdc2, or cyclin B1 levels are observed (∼20% reduction of p-cdc2/cdc2; data not shown), suggesting an alternative mechanism for APRIL-induced G₂/M cell cycle growth arrest.

TRAF/NF-κB activation was reported to mediate the effects of APRIL on transcription, proliferation, survival, and differentiation (34). However, in our system, incubation with APRIL reveals no induction of NF-κB activity (Fig. 3A), in contrast to TNF-α (positive control). We have therefore investigated major intracellular signaling pathways (Akt, p38K, ERK, and JNK) as possible mediators of APRIL effects, as described previously (11, 34). Our data show that APRIL induces JNK and Akt phosphorylation, whereas it has no effect on ERK and p38 (Fig. 3B).

We have verified that this effect was BCMA-dependent, because in anti-BCMA shRNA-transfected cells APRIL was ineffective in inducing phosphorylation of either Akt or JNK molecules (Supplemental Fig. 2). We then focused on the role of Akt and JNK as possible mediators of APRIL antiproliferative action. Incubation of cells with APRIL and the PI3K/Akt inhibitor LY-294002 (40 μM; Fig. 3C) or wortmannin (1 μM; Supplemental Fig. 3A) results in no significant modification of the ligand effect on cell growth, suggesting that neither PI3K nor Akt mediates BCMA-dependent inhibition of HepG2/Hep3B proliferation. In contrast, SP600125 (a pan-JNK inhibitor, 10 μM) (46), had a significant antiproliferative effect of its own (Supplemental Fig. 3B), but no further reduction of cellular proliferation by APRIL was observed.

In view of our finding of increased JNK phosphorylation by APRIL and considering the fact that JNK1 has been demonstrated to be only a positive regulator of proliferation of HCC cells (47), we hypothesized that JNK2 might be the molecule regulating the antiproliferative effect of APRIL. Such an effect has been reported previously in murine embryonic fibroblasts, 3T3 fibroblasts, and erythroblasts (48), in which Jnk2^−/− cells were found to exhibit a proliferative advantage, compared with their wild-type counterparts. To test our hypothesis, we separately knocked down each JNK isoform, with the use of shRNAs against JNK1 and JNK2. JNK1 inhibition led to a massive reduction of cell proliferation, comparable to the effect of SP600125 (Supplemental Fig. 3C), whereas APRIL had no additional effect. In contrast, JNK2 inhibition significantly attenuated the antiproliferative effect of APRIL in HepG2 cells (Fig. 3D), suggesting that JNK2 is implicated in the BCMA-mediated antiproliferative effect of APRIL in HCC cells.

To further decipher the mechanisms underlying these findings, we performed a kinetic transcriptome analysis, after 2, 6, and 12 h of APRIL incubation in HepG2 cells (Fig. 4A), resulting in a time-dependent modification of up- and downregulated transcripts. Gene set enrichment analysis (39) of time-modified genes reveals one significantly inhibited pathway related to immune signaling and nine induced pathways related to the cell cycle and DNA replication (Fig. 4B). Leading edge analysis (39) (Fig. 4B) of the amplified pathways revealed a number of common transcripts implicated in all pathways, a result verified by qRT-PCR (Fig. 4C). They include the following: 1) Minichromosome maintenance (MCM) 2, 4–6, and CDC6. MCM proteins participate in DNA unfolding, recruitment of other proteins, and DNA doubling; their activity is decreased by their phosphorylation by CDC6 (49). These transcripts are decreased specifically after APRIL incubation. 2) PCNA, a cofactor of the DNA polymerase δ complex downstream of the MCM complex, as well as a cofactor of DNA polymerase ε upstream of the MCM complex (50), is modified in seven of nine processes. 3) RPA3 (modified in seven of nine processes) is involved in the S phase of the cell cycle (50). Finally, 4) DNA polymerase ε itself (POLE2) is involved in six of nine processes. The effect of APRIL on these genes could be blocked with the use of shRNA against BCMA (Fig. 4C). These elements, therefore, might represent the primary target of the transcriptional effect of APRIL on the genome via BCMA, explaining the effect of the ligand on the cell cycle.

An alternative analysis of transcriptome data permits the reverse in silico identification of transcription factors potentially modified by an agent (40). In this analysis, by applying the Web resource TFactS (http://www.tfacts.org), we identified only two transcription factors significantly modified by APRIL: FOXO3A (positive association, p = 0.0068) and E2F (negative association, p = 0.0019). Of interest, the regulatory/promoter regions of all identified genes (MCM 2, 4–6, CDC6, PCNA, RPA3, and POLE2) contain putative FOXO3 binding sites, as detected with the online resource multiTF (http://multitf.dcode.org).

To further validate this result, we transfected cells with constructs of FOXO3A or E2F response elements, cloned upstream of the luciferase gene. Unfortunately, both cell lines display very low E2F activity, not permitting analysis of further inhibition. In contrast, FOXO3A transcriptional activity is dose-dependently increased by the addition of APRIL and inhibited by shRNA against BCMA (Fig. 5A, Supplemental Fig. 3D).

Previous reports have shown that Akt induces specific site phosphorylation of FOXO3, reducing its activity (51, 52), in contrast to an activating effect of JNK (52). As expected, addition of the JNK inhibitor SP600125 in APRIL-treated cells resulted in inhibiting APRIL-induced FOXO3 activation in HCC cells (Supplemental Fig. 3D), suggesting that FOXO3 activation is downstream of BCMA-induced JNK phosphorylation; in contrast, the PI3K inhibitor LY-294002 had no effect (Fig. 5A). Furthermore, FOXO3A activity was not modified by the transfection of cells with shRNA against JNK1, whereas shRNA against JNK2 reverted APRIL-induced transcriptional activity of FOXO3A (Fig. 5B). These data suggest that FOXO3A activation is a downstream effector of APRIL/BCMA-induced JNK2 phosphorylation.

A recent report indicates that JNK could directly phosphorylate Akt (53), whereas an interplay between Akt and JNK2 has been reported previously in neuronal cells (54). To study whether Akt activation relies on JNK phosphorylation, we assayed JNK and Akt phosphorylation, after APRIL stimulation, in the absence (control) or presence of the JNK inhibitor SP600125. Our results (Fig. 5C) suggest that inhibition of JNK by SP600125 results in a substantial inhibition of Akt. To further verify our results, we transfected cells with shRNA against JNK1 and JNK2 and studied Akt phosphorylation after APRIL stimulation (Fig. 6). shRNA against JNK2 (but not against JNK1) significantly reduced both APRIL-induced and basal Akt phosphorylation. Furthermore, JNK2 inhibition led to a more intense loss of the 54-kDa isoforms of JNK1/2, whereas JNK1 inhibition had a more prominent effect on the 45-kDa isoforms, in line with previous results in 3T3 fibroblasts (47).

Finally, we report that APRIL enhanced GADD45 mRNA transcription in a time-dependent manner (Figs. 4C, 5D). The increase of GADD45 mRNA transcription induced by APRIL was almost abolished in cells transfected with anti-BCMA shRNA, relaying its transcriptional enhancement to APRIL/BCMA stimulation. GADD45 is a DNA damage-inducible gene, inducing G₂ phase growth arrest via cdc2 inhibition (55). However, as reported above, we could identify only minor changes in cdc2 phosphorylation (∼20% reduction of the p-cdc2/cdc2 ratio in APRIL-treated cells). Interestingly enough, in a very recent study, JNK2 has also been reported to phosphorylate Bcl-xL, which binds and colocalizes with cdc2, leading to G₂ cell cycle arrest (56), whereas GADD45 is also directly induced by FOXO3 expression, through JNK activation (57). GADD45 activation by APRIL binding to BCMA suggests that G₂/M growth arrest might be regulated by one of these mechanisms.

In an attempt to identify whether the APRIL/BCMA system might have a potential implication in HCC, we assayed by immunocytochemistry the expression of APRIL, BAFF, and their receptors in 12 HCC cases and 7 normal livers (Fig. 7). In normal liver specimens, APRIL is detected only in hepatocytes; in contrast, BAFF stains homogeneously the hepatic parenchyma. TACI stains rare occasional hepatic parenchymal cells, whereas BCMA exhibits moderate intracellular hepatocyte immunoreactivity; BAFFR stained preferentially hepatic blood vessels. This profile is modified in HCC: Staining of both APRIL and BCMA shows remarkable enhancement, with all epithelial cells heavily stained. BCMA presents a discrete peripheral staining, suggestive of membrane localization. In contrast, BAFF staining is not modified, whereas BAFFR staining is reduced, concentrated in perivascular cells, and TACI staining is absent. It should be noted that the profile of APRIL/BAFF and their receptors in the HCC samples studied is comparable to that of HepG2 and Hep3B cells, presented in Fig. 1.

APRIL, BAFF, and their receptors (BCMA, TACI, BAFFR) for a long time were considered to be specifically involved in B lymphocyte proliferation and differentiation (58–61). Abnormal expression of APRIL and BAFF has been found to regulate malignant B lymphocyte proliferation, suggesting a role for these ligands in the growth of neoplastic cells (43, 44). However, many authors have reported the expression of APRIL and BAFF in an array of normal and neoplastic cell lines (1, 18, 27–33) and tissues (20–26, 36). The discovery of BCMA and TACI expression in normal tissues other than B lymphocytes (20–24, 37) suggests that these two receptors may also have a function outside the immune system. In this report, we present evidence that APRIL and BCMA are expressed in normal liver tissue and massively induced in HCC specimens and derived cell lines, whereas BAFF and BAFFR are also expressed, but not modified, and TACI is absent. We used HepG2 and Hep3B cells as an in vitro model to assay the exclusive role of APRIL and BCMA. These cells represent a rare system in which BAFF (which is phenotypically ineffective) and APRIL (inducing cell growth arrest exclusively through BCMA) present divergent results, whereas low secretion of BAFF, expressing a 5- to 100-fold lower affinity for BCMA (11–13), compared with that of APRIL, makes improbable its interaction with BCMA, in this model.

In most cases, the effects of ligands acting on BCMA are indistinguishable from those of TACI, which is usually equally present in the studied systems. In previous work, APRIL was reported to stimulate the proliferation of several nonimmune neoplastic cells (19, 26, 62). This effect was attributed to stimulation of the cell cycle, through cyclin D (63, 64), whereas different signaling pathways have been reported in engineered cells with an overexpression of the receptors involving ERK or JNK (34, 65), in addition to the canonical NF-κB activation (66, 67). As reported in this article, in nonengineered HCC cells, APRIL did not modify NF-κB activity; hence, we have assayed the main intracellular pathways, previously described to be induced upon BCMA activation. We report that p38 and ERK are not modified, whereas JNK and Akt were significantly activated by APRIL; however, PI3K inhibition did not reverse this effect. In contrast, JNK1 inhibition led to decreased proliferation, as it was previously reported (47); however, the APRIL antiproliferative effect is not mediated by JNK1. We therefore focused on JNK2, and we report that JNK2 mediates the decrease of cell growth induced by APRIL. Of interest, this differential effect of JNK1 and JNK2 has been reported previously in mouse embryonic and 3T3 fibroblasts and erythroblasts (48, 68). In a very elegant study, Sabapathy et al. (48) showed that JNK1 was proproliferative via c-Jun phosphorylation, whereas JNK2 was antiproliferative through negative regulation of c-jun expression/phosphorylation and ubiquitin binding to c-jun. Furthermore, in a tissue type- and cell type-specific manner, JNK2 has an important role in regulating several molecules related to cell cycle progression checkpoints, especially under DNA damage conditions, leading to cell cycle arrest (69–72). Hepatocytes from Jnk2^−/− mice, however, exhibit increased proliferation rates compared with their wild-type counterparts (73), a finding in line with our results in HCC cells. We further show that APRIL, through BCMA, induces Akt phosphorylation, in a JNK2-dependent manner, suggesting that JNKs may modify the activity of the PI3K pathway, a finding that has been only recently reported (74). Although previous reports came to this conclusion with the use of JNK inhibitors like SP600125, whose specificity has been questioned recently (46), our knockdown experiments clearly confirm this interplay between the two pathways, establishing Akt phosphorylation as a downstream effect of JNK2 activation.

Our observations support the pivotal role of the new pathway reported in this article (APRIL/BCMA/JNK2/FOXO3A/GADD45) in the negative regulation of HCC cellular proliferation. The FOXO family of transcription factors gained increased interest as pivotal elements of cell fate, implicated in diverse functions, from development, longevity and aging, to control of cell survival or apoptosis (75). Some members are directly implicated in cell cycle control, through G₁-related p130 and cyclin G2, or GADD45 resulting in G₂ cycle arrest (76), as equally shown in this study. Of interest, APRIL-regulated transcripts (MCM 2, 4–6, and CDC6), coding for proteins participating in DNA unfolding and doubling, have in their promoter region putative FOXO3A binding sites. GADD45 also participates in G₂ cell cycle arrest, via cdc2 inhibition (77), and its expression has been reported previously to be JNK1- and JNK2-dependent (78). However, in the current study, we could identify only minor changes in cdc2 phosphorylation (∼20% reduction of the p-cdc2/cdc2 ratio in APRIL-treated cells). Interestingly enough, in a very recent work, JNK2 has also been reported to phosphorylate Bcl-xL, which binds and colocalizes with cdc2, leading to G₂ cell cycle arrest (56).

Targeted therapy represents a significant advance in HCC management. Different targets are actually under investigation, including members of the TNF superfamily (reviewed in Ref. 79). Data reported in this article show that APRIL/BCMA might also have potential diagnostic or therapeutic significance. However, an apparent contradiction derives from our findings: We show that the ligand decreases cell growth, whereas it is massively expressed in a highly invasive tumor. Two hypotheses could explain this finding: 1) The space limits of the tumor may provoke a cellular signaling “switch,” providing a local APRIL/BCMA “brake” for further tumor expansion, a result previously reported for other cardinal molecules (80); and 2) in the tumor environment, APRIL, besides regulating the cell cycle, may also act as a proinflammatory regulator of hepatocytes, as we have recently reported in keratinocytes (37). Of interest, a recent report suggests that FOXO3, activated equally by JNK, is responsible for the maintenance of acute myeloid leukemia, initiating cells in their immature state, thereby promoting maintenance of the disease (81). If a similar mechanism is also active in HCC cells, it could equally explain the increased levels of FOXO3-promoting APRIL in HCC specimens. These novel elements of APRIL effects could reorient our understanding of APRIL in HCC tumor biology.

In conclusion, in this study we report a novel signaling pathway for APRIL/BCMA via JNK2 and FOXO3A that regulates HCC cellular proliferation. Furthermore, we report that interplay exists between the MAPK pathway and the PI3K pathway whose importance should be further studied. Finally, our data suggest that BCMA, BAFFR, and TACI may also play an important role in non–immune-related tissues and they can therefore be considered TNF superfamily receptors not unique to B cell-related functions, but with a much broader implication in human physiology and disease.

The authors have no financial conflicts of interest.