Downregulation of JUNB impairs cell cycle progression from G1 to S
To address the role of JUNB in cell proliferation, we investigated the consequences of JUNB depletion on the distribution of cells in the various phases of the division cycle using U2OS cells which are one of the most widely used model for cell cycle studies and that we have previously used to study the regulation of JUNB protein levels during the G2/M transition [11, 13].
First, we searched for reliable anti-JunB siRNAs. Three different siRNAs targeting various regions of the JUNB mRNA (siJUNB-792, siJUNB-803 and siJUNB-848) were designed. As a negative control, we resorted to a commercial control siRNA (siControl). All three JUNB siRNAs strongly downregulated JUNB expression at both the mRNA and the protein level, as assayed 72 h post-transfection (Fig. 1 A, B). RT-qPCR revealed that siJUNB-792 RNA displayed the highest silencing efficiency, leading to an ∼85% decrease in JUNB mRNA steady-state level (Fig. 1A). We also investigated whether the anti-JUNB siRNAs could perturb the abundances of the other JUN family proteins (JUN and JUND). Immunoblot analyses showed that siJUNB-803 affected significantly JUN and JUND levels (Fig. 1B and Additional file 1: Fig. S1A). Therefore, all further experiments were conducted with siJUNB-792 and/or siJUNB-848. In parallel, immunofluorescence assays indicated a homogeneous nuclear decrease in JUNB protein levels after transfection of U2OS cells with siJUNB-792 or siJUNB-848 (Fig. 1C).
Next, U2OS cells were subjected to RNAi for 72 h and their distribution in the cell cycle was analyzed by flow cytometry using propidium iodide-labelling of DNA, while DNA synthesis was assayed in parallel by EdU incorporation. As compared to control conditions, the cells transfected with siJUNB-792, as well as those transfected with siJUNB-848, showed an accumulation in G0/G1 associated with a strong decrease in S phase (Fig. 1 D, E). DNA synthesis was also found reduced in siJUNB-treated cells (Fig. 1D). These observations consequently pointed to a crucial role for JUNB in promoting the progression from G1 to S. Both the decrease in the fraction of cells in S and the G0/G1 arrest were also observed in epithelial cell lines derived from a uterine cervix carcinoma (HeLa), a lung adenocarcinoma (H1395), and a breast cancer (MCF-7) (Additional file 1: Fig. S1B and S1C). Furthermore, cell proliferation assays indicated 25 and 10% reductions in the number of proliferating U2OS cells 72 h after transfection with siJUNB-792 or siJUNB-848, respectively (Fig. 1F, left panel) and slower proliferation rate was confirmed by quantification of living cell time-lapse imaging (Additional file 1: Fig. S1D left panel). On the contrary, overexpression of JUNB obtained by transfection of a eukaryotic JUNB expression vector (pCDNA3-JUNB) for 48 h resulted in a 25% increase in the number of proliferating cells (Fig. 1F, right panel). Higher proliferation rate was confirmed in established U2OS-UTA6 and MCF-7 cell lines overexpressing JUNB (Additional file 1: Fig. S1D middle and right panel). Additionally, to formally establish that the changes in cell cycle arrest observed in RNAi experiments were specific to JUNB downregulation, and not caused by siRNA off-target effects, we performed JUNB rescue experiments. For this aim, U2OS cells were transfected with pCDNA3-JUNB, which expresses a JUNB mRNA insensitive to siJUNB-792, to maintain substantial JUNB protein levels in siJUNB-treated cells and their cell cycle distribution was analyzed 48 h later. Overexpression of JUNB compensated the RNAi-mediated loss of endogenous JUNB (Additional file 1: Fig. S1E, left upper panel) and cell accumulation in G0/G1 was significantly reduced (Additional file 1: Fig. S1E, left lower panel and right panel), indicating that the siJUNB-induced cell cycle arrest was indeed caused by the downregulation of JUNB.
Thus, our data indicate that JUNB exerts a positive action on cell proliferation in U2OS cells and in several other epithelial cancer types, in particular by promoting cell cycle progression from G1 to S.
JUNB-regulated transcriptome
Next, to gain insights into the transcriptional network regulated by JUNB, U2OS cells were transfected with siJUNB-792, siJUNB-848, or siControl for 72 h before RNA preparation and probing of Affymetrix GeneChip Human Transcriptome 2.0 Arrays (containing mainly probes for protein-coding genes) in 3 independent experiments. In total, 1842 and 1475 transcripts were differentially up- or downregulated by a 1.25- to 10-fold factor with an FDR < 0.05 in siJUNB-transfected cells as compared to control cells (Additional file 1: Fig. S2A, Additional file 2: Table S1). The differences between the results obtained with siJUNB-792 and siJUNB-848 may be due to the fact that the former is more efficient at degrading JUNB mRNA, leading to stronger and/or faster downregulation of the JUNB protein (Fig. 1A). However, we cannot formally exclude that part of these differences might also be contributed by differences in some side-effects of the two JUNB siRNAs. Then, for siControl versus each siJUNB comparison, gene set enrichment analyses (GSEA) [33] based on hallmarks pathways were performed. The highest enrichment values were for genes involved in the following pathways (i) G2/M checkpoint and progression through the cell division cycle, (ii) EMT, (iii) TGFβ, (iv) KRAS, (v) E2F, (vi) AKT, (vii) IL2, (viii) angiogenesis, and (ix) response to ultraviolet radiation (Fig. 2A), pointing to likely pleiotropic effects of JUNB in U2OS cells. The fact that the same pathways are found downregulated with both JUNB siRNAs suggests that differences between the two mRNA silencers are more likely due to unequal silencing levels rather than to off-target effects.
Among the genes whose expression varied upon JUNB removal, 111 were upregulated and 135 were downregulated by the two JUNB siRNAs. They were called high-confidence JUNB-regulated genes (Fig. 2B, Additional file 1: Fig. S2A and Additional file 2: Table S1). Consistently with the cell cycle arrest observed upon JUNB depletion in U2OS cells (Fig. 1D), critical cell cycle regulators were found among them (Additional file 3: Table S2). In particular, CCNE1, which codes for cyclin E1 (CCNE1), was one of the most downregulated genes after JUNB depletion, and TGFB2, which codes for the cytokine TGF-β2 (TGFB2), was one of the most upregulated. Both were chosen for functional studies because the former is a crucial positive regulator of the G1/S transition and the latter a negative regulator of cell division [35, 36] (see below for more details). We also observed downregulation of other cell cycle regulators such as DNA damage repair genes (ERCC2), cytokines (IL8), growth factors (FGF5), and downregulation of frizzled class receptor 3 (FZD3), and upregulation of other genes such as GAS2L1 and RPTOR (Additional file 3:Table S2). To validate our Affymetrix array transcriptomic results, we monitored the mRNA levels of these eight genes after JUNB depletion from U2OS cells in independent RT-qPCR assays (Fig. 2C). We also extended our observations to HeLa cells by analyzing CCNE1, TGFB2, IL8, GAS2L1, and RPTOR gene expression by RT-qPCR (Additional file 1: Fig. S2B). Additionally, we confirmed downregulation of CCNE1 and upregulation of TGFB2 upon JUNB silencing in H1395 and MCF-7 cells by RT-qPCR (Additional file 1: Fig. S2C and S2D). This indicated that the regulation of these genes by JUNB is conserved in epithelial cancer cells other than osteosarcoma U2OS cells.
During the G1/S transition, the kinase CDK2 in complex with basal levels of CCNE1 phosphorylates the retinoblastoma protein (pRB), leading to the activation of the transcription factors E2F1, E2F2, and E2F3, a subclass of E2F transcription factor family. In turn, these activated E2Fs stimulates the transcription of CCNE1 and other genes encoding proteins that either promote the progression through G1/S or are required for DNA replication [35], thus creating a positive response loop that strengthens E2F activity and facilitates cell cycle progression [37]. Downregulation of CCNE1 has been shown to arrest cells in G1/S [38], to decrease pRB phosphorylation and to inhibit the E2F signaling pathway [39, 40]. Consistently with the latter observation, as well as with the G1/S cell cycle arrest undergone by JUNB-depleted U2OS cells (Fig. 1D), immunoblotting experiments indicated decreased levels of both CCNE1 and the phosphorylated form of pRB after RNAi-mediated JUNB downregulation (Fig. 2D). Similar results were observed in siJUNB-transfected MCF-7 cells (Additional file 1: Fig. S2E).
Of note, JUNB gene transcription has been reported to depend on E2F1, E2F2, and E2F3 (E2F1-3) activities [35]. As JUNB protein is required for CCNE1 expression (Additional file 2: Table S1), which in turn controls E2F1-3 activities, JUNB downregulation is very likely to lead to cell cycle inhibition by interfering with E2F signaling (Fig. 2A). Supporting this hypothesis, our transcriptomic data showed that JUNB and E2F1-3 share a number of common target genes involved in cell cycle regulation, one of the shared downregulated genes being TGFB2 (Additional file 3: Table S2 and ref. [35]). As TGFB2 is known to be a negative regulator of cell division, the repression of its gene by JUNB may prevent its cytostatic effect. Indeed, we confirmed the upregulation of TGFB2 after JUNB depletion concomitantly with the downregulation of CCNE1 in U2OS and in MCF-7 cells (Fig. 2D and S2E; see functional evidence below). Taken together, these results suggested that JUNB depletion leads to cell cycle arrest in G1/S and impairs progression through S, at least in part, by interfering with the E2F signaling pathway (Fig. 2E).
We also found that several JUNB-regulated genes are involved in both TGFB signaling and EMT (Additional file 4: Table S3) with some of them also being regulated by E2F1-3 (TGFB2, RBPJ, JAG1, CITED2) [35, 41].
In conclusion, our transcriptomic data suggest that JUNB cooperates with and/or is a downstream effector of E2F, TGFB, AKT, and KRAS signaling and not only controls cell proliferation, but may also regulate cellular functions linked to EMT, cell invasion, and metastasis. These data also indicate that changes in JUNB levels can disturb important cell functions, such as the G1/S transition, by altering the expression of its target genes.
Characterization of the JUNB cistrome in U2OS reveals potential direct target genes of JUNB involved in the regulation of cell cycle progression and EMT
As a first step to determine which of the genes whose expression varied in our transcriptomic study could be direct transcriptional targets of JUNB, we next mapped JUNB binding sites genome-wide in U2OS cells by ChIP-seq. Fragmented chromatin was immunoprecipitated with a highly specific anti-human JUNB antibody and immunoprecipitated DNA fragments were sequenced in parallel to input DNA from the same cells. Two independent biological replicates were performed to ensure that the enriched JUNB binding sites were not artifacts of sample processing and/or sequencing. The MACS peak-calling algorithm was applied to both replicates using the parameters indicated in “Materials and methods”. This led to the identification of 7572 JUNB binding peaks common to both replicates (Fig. 3A). Among these peaks, JUNB binding was more frequently observed in intergenic (39%) and intronic (42%) regions than in transcriptional promoters (defined as the 10 kb lying upstream of transcription start sites (TSSs)) of the nearest gene (11%). Moreover, JUNB binding sites were poorly represented in exons (2%), as well as in the 10-kb regions downstream of the transcription termination sites (TTSs) of the nearest gene (6%) (Fig. 3B). The distance distribution between JUNB binding sites and the closest gene TSSs, regardless of whether the former was located upstream or downstream of the TSS, is represented in Fig. 3C. About 62% of JUNB binding sites (4748/7572) were found located within 50 kb of the nearest TSSs. Our data were consistent with the currently emerging notion that AP-1-binding sites usually do not reside within gene promoters [1, 42].
We then addressed the distribution of JUNB binding sites between enhancers and promoters as described in [42]. To this aim, we intersected our JUNB cistrome data with publicly available genome-wide data obtained in U2OS cells for three histone modifications, H3K4me1, H3K4me3, and H3K27ac [43], classically used to discriminate active or inactive promoters and enhancers [42]. Heatmaps of the three histone modifications showed that the distribution of H3K4me1, H3K4me3, and H3K27ac marks was bimodal around JUNB peak centers (Fig. 3D), which is typical of regulatory elements bound by transcription factors [44]. Among the 7572 JUNB binding peaks, (i) 4257 were not surrounded by any of the three histone marks (62.3%) and were attributed no function, (ii) 2042 were surrounded mainly by H3K27ac and H3K4me1 (30%), which indicated candidate active enhancers (cAE), and (iii) 207 were surrounded by H3K27ac mainly in co-occurrence with H3Kme3 (3%), which indicated candidate active promoters (cAP) (Fig. 3E). Only 0.1% of JUNB binding peaks were found in candidate inactive promoters (cIPs; regions marked by H3K4me3 but not by H3K27ac) and 4.3% in candidate inactive enhancers (cIEs; regions marked by H3K4me1 but with low H3K27ac content and carrying neither cAPs nor cIPs) (Fig. 3E, Additional file 5: Table S4). Importantly, the majority of JunB-binding cAEs (96.4%) and cAPs (92.8%) were situated in open chromatin domains, as deduced from publicly available ATAC-seq data obtained in U2OS cells [45] (Additional file 5: Table S4). This strengthened the idea that they are genuine active regulatory elements.
De novo DNA motif analysis using the MEME suite [46] on all of the 7572 JUNB binding peaks revealed several highly enriched transcription factor (TF)-binding motifs (TFBMs). The TRE binding site showed the highest enrichment by far (Fig. 3F). It should be noted that the CRE motif appeared less enriched than any of the abovementioned TFBMs, suggesting that JUNB has lower affinity for it than for TREs in U2OS cells. Indeed, the CRE motif was found in only 20% of the JUNB peaks, whereas one or several copies of TRE were found in 90% of these peaks. The latter observation also supported the notion of direct binding of JUNB to these sites. Finally, the fact that a variety of TFBMs was found in many JUNB peaks suggested that JUNB may collaborate with other TFs at the enhancers and promoters to which it binds to regulate its target genes.
Thus, our data indicate that, besides regions for which no regulatory function could be attributed, JUNB binds more frequently to candidate active enhancers than to candidate active promoters in U2OS cells. JUNB binding to these sites is very likely direct as 90% of them contain at least one TRE motif. Finally, a majority of JUNB-bound candidate enhancers are located within 50 kb of the nearest TSSs, i.e., in a relative gene vicinity.
To search for possible direct transcriptional target genes of JUNB in U2OS cells, we then crossed our ChIP-seq- and transcriptomic data. Only the JUNB ChIP-seq peaks belonging to cAE, cAP, cIE, and cIP (2582 out of 7572) were considered for this analysis. The identified peaks were assigned by annotation to the closest gene TSS to 2096 putative target genes (Additional file 5: Table S4). Among the high-confidence gene list (246 genes), 26 genes had at least one JUNB peak suggesting that these genes may represent direct JUNB-regulated target genes (Fig. 3G, Additional file 6: Table S5 and Additional file 1: Fig. S3). Several of them (TGFB2, ERCC2, RPTOR, RBPJ, ETV4, CDH4, CORO2B, TSPAN2) being involved in cell cycle regulation and EMT (Additional file 3: Table S2 and Additional file 4: Table S3), strengthening the notion that JUNB may regulate the transcription of genes involved in the regulation of cell cycle and proliferation, TGFB signaling, and EMT. The other high-confidence JUNB-regulated genes identified in the transcriptomic assay without JUNB binding sites could be indirect target genes.
JUNB-dependent regulation of TGFB2 during the G1-to-S transition
Studying in more detail TGFB2 as a candidate direct transcriptional target of JUNB (Additional file 1: Fig. S3 and Additional file 6: Table S5) was of particular interest, not only because it was one of the most upregulated genes upon JUNB RNAi in U2OS cells (Additional file 2: Table S1), but also because its protein product has formerly been reported to inhibit proliferation of carcinoma cells [32] and to promote both cell invasion and cancer progression [47, 48].
Our ChIP-Seq data indicated 2 JUNB peaks located in open chromatin domains defined by ATAC-seq at the TGFB2 locus associated with histone marks. One was located at +31 kb. It contained both a TRE and a CRE motif and fell in a cIE. The other JUNB peak lied at +114 kb. It contained a single TRE motif and fell in a cAE (Additional file 6: Table S5, Fig. 4A and Additional file 1: Fig. S4A). Unless the +31 kb JUNB-bound cIE has no effect on TGFB2 expression, this raised the possibility that the transcriptional regulation of TGFB2 might involve a delicate balance between negative and positive regulatory elements.
ChIP-qPCR were conducted to validate the binding of JUNB to these two TGFB2 gene most proximal sites (Fig. 4A). To this aim, as well as to further study the effect of JUNB on TGFB2 expression, we resorted to a U2OS tetracycline (Tet)-off cell system, expressing either inducible JUNB (UTA6-JUNB) or empty vector (UTA6-Control) we formerly engineered [11, 13]. Binding of JUNB at +31 and +114 kb was observed in both cell lines, when grown asynchronously in the absence of Tet, nevertheless with stronger signals in UTA6-JUNB cells where the JUNB gene was overexpressed (Fig. 4B). Concomitantly to higher binding of JUNB, lower TGFB2 mRNA and protein levels were observed in JUNB-overexpressing cells (Additional file 1: Fig. S4B and S4C). Lower TGFB2 levels were also observed in MCF-7 cells overexpressing JUNB-GFP (Additional file 1: Fig. S4D). In contrast, higher levels were observed in siJUNB-transfected MCF-7 cells (Additional file 1: Fig. S2E). Together with our JUNB transcriptomic data, this strengthened the idea of TGFB2 being a directly repressed transcriptional target of JUNB.
Next, we addressed whether JUNB could act as a repressor of TGFB2 specifically during the G1-to-S transition and whether this could contribute to cell proliferation regulation. To answer this question, we first arrested UTA6-control and UTA6-JUNB cells in mitosis and, after release in the cell cycle, analyzed both progression through the cell cycle and TGFB2 expression. UTA6-JUNB cells overexpressing JUNB were found to reach S phase faster than cells expressing endogenous levels of JUNB (UTA6-Control) (Fig. 4C and Additional file 1: Fig. S4E; see time points 9 to 15 h). This result was consistent with the higher proliferation rate of asynchronously growing UTA6-JUNB cells overexpressing JUNB in the proliferation assays presented in Additional file 1: Fig. S1D (middle panel). Moreover, TGFB2 mRNA and protein levels were found lower in JUNB-overexpressing cells than in control cells at all times analyzed (Fig. 4 D, E). Of note, TGFB2 mRNA and protein levels were lower in M-arrested UTA6-JUNB than in UTA6-Control cells. We interpreted this observation to result from the higher basal level of JUNB at time 0 in the former cells (Fig. 4E), due to the cell synchronization- and JUNB expression induction procedures used. Also remarkable, the extent of protein-level reduction (Fig. 4E) was much stronger than that of its mRNA (Fig. 4D) in UTA6-JUNB cells, suggesting that JUNB can alter TGFB2 protein production, not only via a direct effect on its encoding gene, but also indirectly via influencing post-transcriptional mechanisms involved in TGFB2 production (see below for more details). It is also interesting to underline that, at time 15 h, when a large fraction of UTA6-JUNB cells have already entered the S phase (Fig. 4C), a slight but detectable increase of TGFB2 was detected at both the RNA (Fig. 4D) and the protein level (Fig. 4E). As no such an increase was detectable in control cells (which were still essentially in G1 and early S at the same time point), this supported the idea that TGFB2 repression by JUNB largely occurs during the G1-to-S transition. Further strengthening this notion, (i) JUNB binding at +31 and +114 kb could be confirmed in both UTA6-JUNB and UTA6-Control cells during the G1/S transition, being the highest binding enrichment at +31 kb in all cases (Fig. 4F) with, however, (ii) a stronger binding in UTA6-JUNB cells (Fig. 4F); and (iii) higher JunB binding enrichment was found at +31 and +114 kb in G1-to-S-synchronized than in exponentially growing cells (Fig. 4 B, F).
Finally, we asked whether the inhibition of TGFB2 production by JUNB was important to allow cells to progress from G1 to S. To this aim, we tested whether the addition of exogenous TGFB2 could affect progression from M to S of nocodazole-synchronized UTA6-Control- and UTA6-JUNB cells. Such a treatment (i) impaired G1-to-S progression in control cells (Fig. 4G), consistently with the well-described cytostatic effect of TGFB2, and (ii), more interestingly, slowed down the accelerated M-to-S progression of UTA6-JUNB where JUNB was induced (Fig. 4G). The latter slowdown was evidenced by (i) the higher number of cells still in G0/G1 and (ii) the lower number of those in S phase seen in TGFB2-treated- versus non-treated UT6-JUNB cells both at 12 and 15 h after cells were released in the cell cycle, consistently with the cytostatic effect of TGFB2 [11].
Although we cannot discard that overexpression of JUNB may alter AP-1 dimer composition and have a global impact on gene expression and affect cell cycle progression, our data indicate that the increasing expression of JUNB during progression from G1 to S represses the expression of TGFB2, which would otherwise behave as a cell cycle brake. Additionally, our data support the idea that JUNB-mediated downregulation of TGFB2 synthesis implies both transcriptional and post-transcriptional mechanisms.
JUNB-dependent regulation of CCNE1 during progression from G1 to S
The finding that CCNE1 was one of the most downregulated high-confidence genes upon JUNB knockdown in our JUNB transcriptomic experiments (Additional file 2: Table S1) was interesting, due to the acknowledged role of its protein product in cell division control. We therefore formally asked whether this gene was another target whereby JUNB would actually control cell cycle progression. This was achieved in several steps. First, asynchronous U2OS cells were subjected or not to JUNB RNAi transfection and pCDNA3-JUNB (Fig. 5A), and CCNE1 protein expression were monitored. In the absence of pCDNA3-JUNB, the decrease in CCNE1 paralleled that of JUNB whereas, upon (partial) JUNB rescue, this reduction was more limited. Similar results were observed in MCF-7 cells transfected with siJUNB and pCDNA3-JUNB (Additional file 1: Fig. S5A). This strengthened the idea of CCNE1 gene expression being dependent on JUNB. Second, an ectopic CCNE1 protein was expressed by transfection of pCDNA3-CCNE1 plasmid into JUNB-depleted U2OS cells (Fig. 5B). This not only limited the disappearance of CCNE1, but also partially restored the ability of these cells to exit the G1 arrest (Fig. 5C), pointing to JUNB-dependent regulation of CCNE1 as important for cell cycle control. Finally, as CCNE1 transcription is known to be tightly regulated during the G1-to-S transition [49], CCNE1 mRNA (Fig. 5D) and protein (Fig. 4E) levels were assayed in UTA6-JUNB- and UTA6-Control cells released in the cell cycle from a nocodazole block (Fig. 4C). Both CCNE1 mRNA (Fig. 5D) and protein (Fig. 4E) abundances increased in the two cell contexts during progression towards S, but more strongly in JUNB-overexpressing UTA6-JUNB- than in control cells. Moreover, increased CCNE1 mRNA and protein expression occurred earlier in the former- than in the latter cells and largely paralleled JUNB levels in both cases. This provided further evidence for an important role of JUNB-dependent regulation of CCNE1 expression in the progression of G1 cells towards S.
CCNE1 was not identified as an obvious possible direct candidate JUNB target gene in our JUNB cistrome analysis. Nevertheless, we noted the presence of two TREs located 6.4 and 6.1 kb upstream of the CCNE1 TSS. Interestingly, this region overlaps with H3K4me1 and H3K27ac marks and lies in an open chromatin region in U2OS cells, suggesting it to be an active enhancer region (Additional file 1: Fig. S5B). As CCNE1 gene transcription is restricted to the limited period of the G1-to-S transition [49], we hypothesized that JUNB binding was not detected in our ChIP-seq experiments because G1/S cells represented only a minor fraction of the asynchronously growing U2OS cell population. Consequently, to address the possibility of G1/S-restricted binding of JUNB to the −6.4/−6.1-kb region, we resorted to synchronized UTA6-Control- and UTA6-JUNB cells released in the cell cycle from a nocodazole block and conducted ChIP-qPCR assays using three sets of primers. Two sets amplified the region containing the −6.4 (A) and −6.1 kb (B) TREs and one amplified a TRE-devoid domain (C) taken as a negative control (Fig. 5E). JUNB binding turned out to be significantly enriched at −6.1 kb TRE site in both cell lines during G1 (time point 3 h) and G1/S (time point 9 h), and, as expected, enrichment levels were stronger in UTA6-JUNB- than in UTA6-Control cells due to the higher levels of JUNB (Fig. 5F).
Thus, taken together, our data indicate that JUNB-dependent G1/S-restricted regulation of the CCNE1 protein is crucial for cells to progress from G1 to S, most probably via direct transcriptional stimulation of this gene by JUNB.
Long-term stimulation of JUNB-overexpressing cells by exogenous TGFB2 promotes EMT and cell invasion
As mentioned earlier, advanced tumors are recurrently bathed in high concentrations of TGFB. Under this condition, TGFB signaling activation in cancerous cells is often associated with, not only a loss of cell proliferation inhibition activity, but also a switch that facilitates EMT, cell migration, invasion and, thereby, promotion of tumor progression and metastasis [50, 51]. The fact that we observed JUNB-dependent repression of TGFB2 in the experiments presented above led us to wonder about the possible disturbance of TGFB2 downstream signaling in JUNB-overexpressing tumor cells growing in a TGFB2-rich environment.
As a first step to address this point, we tested whether increasing the stimulation time by exogenous TGFB2 of JUNB-overexpressing U2OS cells, which would mimic permanent production of this cytokine in an advanced tumor environment, could affect cell proliferation. This turned out to be the case, as the proliferation of UTA6-JUNB cells overexpressing JUNB cultured in the presence of TGFB2 was slowed down at 48 and 72 h (Fig. 6A), in keeping with the data presented in Fig. 4G, but was no longer inhibited in a 96-h-long experiment (Fig. 6A). In contrast, TGFB2 cell proliferation inhibition activity was maintained, at 48, 72, and 96 h (although to a lesser extent), in UTA6-Control cells processed in parallel (Fig. 6A). Thus, compared to control cells, JUNB-overexpressing cells counteract more rapidly/efficiently the inhibitory effect of a long stimulation by TGFB2 on cell growth.
Next, the observation that overexpression of JUNB could contribute to the neutralization of the antiproliferative effect of exogenous TGFB2 upon long-term stimulation led us to ask whether this could be associated with a TGFB2-dependent pro-EMT phenotype. This question was also motivated by, not only the observation by others that JUNB can contribute to EMT in response to TGFB1 in breast and lung cancer cells [30, 31], but also the functional pathway enrichment analysis of our JUNB transcriptomic data pointing to a possible regulatory role for JUNB in TGFB and EMT signaling pathways (Fig. 2F). To address this point, we monitored in UTA6-control and UTA6-JUNB cells cultured for various periods of time in the presence of exogenous TGFB2, the expression of protein markers whose expression is well-known to vary during EMT induction [47] (Fig. 6B). Our data showed that in UTA6-Control cells, the addition of TGFB2 induced the expression of the endogenous JUNB gene transiently (up to 48 h) and in a moderate manner. This induction was associated with concomitant inductions of ZEB1, SMAD2, pSMAD2, SNAI1, fibronectin, and integrin α5. Additionally, slight increases in the mesenchymal markers’ integrin-β1 and vimentin were also observed at the latest time points tested. Interestingly, the induction of all of the tested EMT-associated markers was higher in JUNB-overexpressing than in control cells. It was also of note that the levels of SMAD2, SNA1, fibronectin, integrin-β1, integrin-α5, and vimentin at time 0 were also higher in UTA6-JUNB- than in UTA6-Control cells. It could be plausible that JUNB overexpression may have a global impact on gene expression by altering AP-1 dimer composition thus promoting EMT. The expression of JUNB and EMT markers were also analyzed in established MCF-7 cells expressing JUNB-GFP (MCF-7-JUNB) or GFP (MCF-7-Control) treated with exogenous TGFB2 with similar results (Additional file 1: Fig. S6A). MCF-7 cells are positive for epithelial markers, such as E-cadherin. As expected, TGFB2 treatment induced a higher decrease in the expression of E-cadherin in the MCF7-JUNB cells compared to control cells. These observations supported the idea that high JUNB expression might potentiate exogenous TGFB2-dependent EMT.
Then, we tested whether long exposure to TGFB2 could induce a mesenchymal-like morphotype in UTA-Control- and/or UTA6-JUNB cells. To this aim, filamentous actin staining with phalloidin was used to assess potential morphology changes. Untreated UTA6-Control cells showed cuboidal-shaped and in close contact but, in the presence of TGFB2 for 72 h, they became more loosely arranged and more elongated. In contrast, non-treated JUNB-overexpressing UTA6-JUNB cells showed an elongated shape with less tight connections. Moreover, this phenotype was amplified in the presence of TGFB2 with even less inter-cell connections and an even more elongated morphology associating with visible cytoplasmic extensions (Fig. 6C). This further supported the idea that, in the presence of exogenous TGFB2, JUNB can promote a mesenchymal-like phenotype.
As EMT often correlates with increased invasion ability, we next conducted in vitro invasion assays in Matrigel (Fig. 6D). UTA6-JUNB cells overexpressing JUNB showed more invasive behavior when grown in the presence of exogenous TGFB2 than in its absence. They also showed more invasive than UTA6-Control cells grown in the presence of TGFB2.
Thus, altogether, our data indicated that long-term exposure to exogenous TGFB2, as occurs in a number of advanced tumors, promotes a molecular switch in JUNB-overexpressing cells, which both counteracts the antiproliferative action of this cytokine and promotes EMT at the molecular, morphological, and functional levels.
Regulation of endogenous TGFB2 production by JUNB in the presence of exogenous TGFB2
To clarify the apparent contradiction between JUNB-induced repression of TGFB2 gene, on the one hand (Fig. 4 D, E and Additional file 1: Fig. S4C and S4D), and JUNB acting as a promoter of TGFB2-dependent EMT, on the other hand (Fig. 6 B–D), we hypothesized that, upon long stimulation with exogenous TGFB2, the repressive action of JunB on the endogenous TGFB2 gene could be reduced. In this scenario, exposure of JUNB-overexpressing cells to exogenous TGFB2 should trigger endogenous TGFB2 production which would then promote sustained EMT. To test it, we monitored TGFB2 mRNA levels by RT-qPCR in UTA6-Control- and UTA6-JUNB cells 96 h after addition of exogenous TGFB2 to the culture medium. TGFB2 mRNA accumulation was found to be transiently stimulated with a peak by 48 h post-addition of exogenous TGFB2 in both cell lines. However, TGFB2 mRNA levels were always lower in UTA6-JUNB- than in UTA6-Control cells, including at time 0 (Fig. 6E). This supported the idea that JUNB kept at least some of its transcriptional repression ability on the TGFB2 gene under the conditions studied.
Then, we addressed endogenous TGFB2 protein production. TGFB2 is produced as a precursor preprotein (48 kDa) that is proteolytically processed to give both a so-called latency-associated peptide (LAP) and the mature TGFB2 cytokine. This allows the formation of TGFB2 homodimers (each monomer being 12.5 kDa) bound by disulfide bridges that remain in complex with LAP (and possibly other proteins) until they are secreted outside of cells where they dissociate from LAP and, thereafter, play their biological role [52]. The kinetic analysis of the TGFB2 precursor preprotein by immunoblotting (Fig. 6F) indicated no detectable changes in UTA6-Control cells despite the observed transient induction of the TGFB2 mRNA (Fig. 6E). In contrast, a slight transient increase in the TGFB2 precursor preprotein as well as in the mature monomeric form (Fig. 6F) paralleling the profile of TGFB2 mRNA accumulation (Fig. 6E) was seen in UTA6-JUNB cells. In addition, we assessed endogenous TGFB1 gene and protein expressions according to its activity as an inducer of EMT. In the kinetic analysis of TGFB1 mRNA after exogenous addition of TGFB2 only subtle changes were transiently observed (Additional file 1: Fig. S6B). At the protein level, the kinetics of the TGFB1 precursor preprotein and mature monomeric form between UTA6-Control and UTA6-JUNB cells were similar, although an increase of TGFB1 protein was observed in UTA6-JUNB cells (Additional file 1: Fig. S6C).
The observation of the increment in TGFB2 precursor and mature protein levels were paradoxical, as they contrasted with the lower abundance of the TGFB2 mRNA in UTA6-JUNB cells. This led us to inquire about the possibility of JUNB-regulated translation of the TGFB2 mRNA in the latter cells. To this aim, we quantified TGFB2 mRNA abundance in polysome-enriched microsomal fractions prepared from UTA6-Control- and UTA6-JUNB cells cultured in the absence or in the presence of TGFB2 ligand for 48 h (i.e., the time corresponding to maximal accumulation of TGFB2 mRNA in UTA6-Control and UTA6-JUNB cells; Fig. 6E). No change in TGFB2 mRNA association was seen in the former cells whereas an increase was clearly detected in JUNB-overexpressing UTA6-JUNB cells due to TGFB2 treatment (Fig. 6G).
Thus, sustained exposure of U2OS cells to exogenous TGFB2 reverses the action of JUNB on the production of the endogenous TGFB2 protein from a negative to a positive effect. However, increased production of the endogenous TGFB2 cytokine appears not to be due to loss of ability to repress the synthesis of the TGFB2 mRNA by JUNB. Rather, it results from a TGFB2-induced, JUNB-dependent signaling facilitating TGFB2 mRNA translation via increased recruitment in polysomes (See proposed model of JUNB as a promoter of cell proliferation and cell invasion in Fig. 6H).
Overexpression of JUNB promotes tumor development and metastasis by U2OS cells in vivo
The fact that JUNB overexpression may confer a stronger EMT- and invasion-promoting potential in vitro to JUNB-overexpressing UTA6-JUNB cells as compared to UTA6-Control cells (Fig. 6 B–D) led us to ask whether high JUNB levels could also be associated with higher U2OS cell tumorigenicity in vivo. To this aim, UTA6-JUNB- and UTA6-Control cells were inoculated subcutaneously to immunocompromised NSG mice. Tumor size was then monitored at regular intervals and mice were eventually euthanized at 3 months post-inoculation. Although both cell lines were tumorigenic in all the animals used (5 per group), UTA6-JUNB cell-derived tumors were much bigger at the end of the experiments than UTA6-Control cell-derived ones (Fig. 7A, upper panel) and developed much faster after a shorter latency time (Fig. 7A, lower panel).
We also searched for lung and liver metastases in euthanized xenografted mice. This analysis was facilitated by the GFP-marking of xenografted cells, as the Tet-regulated plasmid used to generate them is a bicistronic vector expressing this fluorescent protein from an IRES sequence [13]. The feasibility of GFP-tracking of cancer cells was established by analyzing primary tumors (Fig. 7B, left upper panel). Whereas neither lung nor liver metastases were found in UTA6-Control cells, some were detected in both organs in 4/5 mice xenografted with UTA6-JUNB cells (Fig. 7B, upper right and lower panel). Hematoxylin and eosin staining (H&E) images of paraffin-embedded sectioned tissue slices from primary tumors and lung and liver metastasis are shown in Additional file 1: Fig. S7A. Specific tumor cells expressing GFP were also detected by immunohistochemistry (IHC) (Additional file 1: Fig. S7B-D) and immunofluorescence (IF) (Additional file 1: Fig. S8) assays. Importantly, IF analysis confirmed that both UTA6-JUNB primary tumors and liver and lung metastasis co-expressed GFP and JUNB protein (Additional file 1: Fig. S8). JUNB expression was observed at all tumor sites tested with signals stronger in UTA6-JUNB cell-derived primary tumors and metastases than in UTA6-Control cell-derived primary tumors analyzed by IF (Fig. 7C) and IHC (Additional file 1: Fig. S9A left panel, S9B and S9C). These data supported the idea that high JUNB expression confers a stronger tumorigenic potential to U2OS cells, leading to greater metastases.
Next, we examined CCNE1 expression by IF and IHC assays and observed that its expression was slightly higher in UTA6-JUNB cell-derived primary tumors than in UTA6-Control cell-derived primary tumors (Fig. 7D and Additional file 1: Fig. S9A left panel). Moreover, CCNE1 expression in UTA6-JUNB cell-derived liver metastases was higher in comparison to UTA6-JUNB cells in primary tumors (Fig. 7D and Additional file 1: Fig. S9B), suggesting that even higher expression of CCNE1 is also associated with metastasis. We also addressed TGFB2 expression in primary tumors and metastases by IF and IHC. No detectable differences in signal intensities could be detected between UTA6-Control and UTA6-JUNB cell-derived primary tumors when examinations were carried out inside tumors (Fig. 7E and Additional file 1: Fig. S9A left panel). However, stronger TGFB2 signals were recurrently found at the border of UTA6-JUNB cell-derived primary tumors (Fig. 7F middle panel and Additional file 1: Fig. S9A left panel), which was also where stronger JUNB signals were often observed (Fig. 7F, left panel, and 7H, and Additional file 1: Fig. S9A left panel). In contrast, CCNE1 protein levels remained similar in the border compared to the central zone (Fig. 7F, right panel, and Additional file 1: Fig. S9A left panel). Of note, no differences in JUNB and TGFB2 expressions were observed between the central and peripheral zones in the UTA6-Control primary tumors (Fig. 7G). We also observed strong fibronectin and integrin β-1 signals at the border UTA6-JUNB cell-derived primary tumors, suggesting that these cells have stronger migration and invasion capacities (Fig. 7 I and Additional file 1: Fig. S9A right panel). Accordingly, TGFB2 expression was stronger in UTA6-JUNB cell-derived liver metastases as compared to UTA6-JUNB cell-derived primary tumors (Fig. 7E, right panel, and S9B), suggesting that the cells with the highest metastatic capacity are also those showing the highest TGFB2 levels.
Thus, taken together, our data indicate that JUNB overexpression in U2OS cells promotes both tumor development and metastasis in vivo. They also suggest that the presence of TGFB2 in the tumor microenvironment may promote cell invasion and the metastatic capabilities of JUNB-overexpressing cells.
High levels of JUNB protein associated with poorer outcomes in epithelial breast cancers
We next addressed whether high JUNB levels could be associated with bad prognosis in cancers by analyzing cancer data sets from the publicly accessible TCGA databank with cBioportal for Cancer genomics [53]. We found that JUNB copy number alterations (CNA) occur in several cancers including sarcoma, ovarian, esophagus, lung squamous carcinoma, and breast cancers (Additional file 1: Fig. S10A). Moreover, breast and ovarian cancer patients with high JUNB levels due to gene amplification showed poorer prognosis than the rest of patients (Additional file 1: Fig. S10B). As the publicly available data on protein expression in osteosarcoma are scarce, we addressed breast cancer, for which a wealth of data can be processed. In silico analysis of available published protein breast cancer dataset [54] using the Kaplan-Meier plots website [55] also showed that high levels of JunB protein were associated with bad prognosis (Additional file 1: Fig. S11A). In addition, JUNB protein levels correlated with cyclin E1 and TGFB2 protein levels (Additional file 1: Fig. S11B), suggesting that high JUNB protein levels may also promote aggressiveness of breast tumors by positively controlling their expression with consequences on both cell proliferation and invasion.