Expression of nc886, PKR, and eIF2α in cholangiocyte and CCA cells
We measured nc886 in a battery of bile duct cells (). Like other non-malignant human cells in our previous studies (1
), nc886 was expressed in immortalized but not malignant cholangiocyte MMNK1 cells (). nc886 was also expressed in three CCA cells (M156, M214, and M055), whereas it was not expressed in M213, its derivative lines M213L0H and M213L5H, and M139 cells. For comparison, all of the bile duct cells constitutively expressed vtRNA1-1, a canonical vtRNA ().
Cell lines used in this study
Expression of nc886, vtRNA1-1, PKR, and eIF2α in cholangiocyte and CCA cells
While PKR was constitutively expressed in cholangiocyte and CCA cells (total PKR in ), it was activated in all the CCA cells except for M156, as indicated by P-PKR Western blot (). Similarly to P-PKR, P-eIF2α was elevated in all the CCA cell lines, but again not in the non-malignant MMNK1 (). In our experiments in , it should be noted that all assays were performed in a normally growing condition without exposure to dsRNA, a natural activator of PKR that leads to apoptosis.
In general, CCA cells tended to have decreased nc886 but elevated P-PKR/P-eIF2α relative to non-malignant MMNK1 cells ( and ). Certainly, there were exceptional cell lines; nc886 was not silenced in all the CCA cells, and nc886 was co-expressed with P-PKR in M214 and M055 cells. This is explained by the genetic diversity of CCA cells. PKR has been known to be regulated by a number of intra-/extra-cellular factors besides dsRNA [reviewed in (6
)]. For example, natural PKR regulators such as PACT or p58IPK may have contributed to the activation of PKR in M214 and M055 cells. We speculate that nc886 is also under complex regulation, as a number of transcription factors are predicted to bind to the vicinity of the nc886 locus.
Taken together with our previous result that suppression of nc886 activates PKR to P-PKR (1
), the anti-correlation between nc886 and P-PKR in a battery of cholangiocyte/CCA cells () was intriguing and raised these questions: 1) is the expression of nc886, P-PKR, and P-eIF2α related causally during tumorigenesis? and 2) how do CCA cells overcome P-PKR/P-eIF2α's apoptotic effect and proliferate? We addressed these issues in the following experiments. As we used several cell lines with different expression of nc886, P-PKR, and P-eIF2α, we designated their expression by +/− signs in a parenthesis, for easy reading. For example, MMNK1 cells are (nc886+, P-PKR−, P-eIF2α-; see also ).
Causal relationship between nc886 and P-PKR
A number of papers have reported PKR activation in malignancies; however, most of them did not elucidate a mechanism for that activation. In one report (18
), a PKR inhibitor protein p58IPK (30
) was suggested to be the cause, because p58IPK was decreased in breast cancer cells whereas PKR was activated. However, this observation was not reproducible in other papers, and thus p58IPK's universal involvement in PKR regulation seemed unlikely (17
). In this study we found that the expression of p58IPK was consistent among the cell lines (), and therefore ruled out its role in PKR regulation in CCA.
nc886 suppressed PKR in CCA cells
Based on our data in and our previous study (1
), we hypothesized that suppression of nc886 activated PKR in CCA cells. To investigate the nc886/PKR relationship, we suppressed nc886 by transfecting antisense oligonucleotides (anti-oligos) into MMNK1, M156, and M214 cells (all nc886+ cells) and then measured P-PKR. Compared to an anti-oligo against vtRNA1-1 (“anti-vt21_2”) as a negative control, anti-oligos against nc886 (“anti886s”) successfully negated nc886 and induced P-PKR in all the three cell lines examined (). Interestingly, there were two distinct mechanisms for anti-oligos to induce P-PKR. In the case of non-malignant MMNK1 cells, the anti-oligos depleted nc886's intracellular level (Northern blot in ). In M156 and M214 CCA cells, the anti-oligo interfered with nc886/PKR interaction (pull-down assay in ).
Our cellular data were reinforced by in vitro experiments. Addition of the anti-oligo against nc886 in MMNK1 lysate (nc886+) increased PKR activity, as measured by in vitro kinase assays (). Conversely, addition of in vitro transcribed synthetic nc886 in M213 lysate (nc886−) inhibited PKR activity (). We also showed their direct physical interaction by electrophoretic mobility shift assays (EMSA) with synthetic nc886 and purified PKR from E. coli ().
Collectively, our data demonstrated that nc886 alone is necessary and sufficient for suppression of PKR through direct physical interaction. nc886 keeps PKR repressed in MMNK1 cells, and PKR is released from that repression when nc886 is suppressed in CCA cells. However, nc886 levels could not explain P-PKR in all the CCA cells, for example, in M214 and M055 (nc886+, P-PKR+). As mentioned earlier, their intrinsic level of P-PKR could have been activated by a factor other than nc886 or p58IPK, or in these malignancies there seems to be a higher threshold level of nc886 necessary for suppression of P-PKR.
The canonical PKR/ eIF2α pathway operates in MMNK1 cells but not in CCA cells
One pre-requisite for our tumor surveillance model is that the canonical PKR/eIF2α pathway operates normally and leads to apoptosis in most cells, because the suppression of nc886 would eliminate such cells through this pathway. We attempted to recapitulate such a situation in non-malignant MMNK1 cells (nc886+, P-PKR−, P-eIF2α-). Upon nc886 depletion, P-PKR phosphorylated eIF2α, decreased global protein synthesis and inhibited cell proliferation ().
nc886 depletion provoked the canonical PKR/eIF2α pathway leading to apoptosis in cholangiocyte MMNK1 cells, but not in CCA cells
P-PKR is known to induce apoptosis mainly through two pathways involving FADD/caspase-8 and APAF/caspase-9, both of which merge onto caspase-3 [reviewed in (6
)]. Consistently, we detected the active form of cleaved caspase-3 and consequent cleavage of PARP (poly(ADP-ribose) polymerase) upon nc886 knockdown in MMNK1 cells (lane 1-2 in ). In contrast, neither caspase-3 nor PARP cleavage was seen in M156 and M214 CCA cells (lane 3-6 in ), although PKR was indeed activated by nc886 depletion (). Hence, there were two distinct consequences of P-PKR activity between MMNK1 cells and CCA cells. It is worth noting that M156 and M214 CCA cells were both P-eIF2α+ ().
To interrogate which step of the PKR pathway is abrogated in these two cell lines, we measured P-eIF2α and global protein synthesis upon nc886 knockdown (). In the case of M156 cells (nc886+, P-PKR−, P-eIF2α+), P-eIF2α was induced but global protein synthesis was not significantly decreased. In the other case of M214 cells (nc886+, P-PKR+, P-eIF2α+), the induction of P-eIF2α was not seen and consistently global protein synthesis was unaffected.
So far, we have shown that the canonical PKR/eIF2α pathway operated normally in non-malignant MMNK1 cells, but not in CCA cells. Phosphorylation of eIF2α (M214) or inhibition of global protein synthesis (M156) malfunctioned so that the two CCA cells escaped from apoptosis upon nc886 suppression.
The PKR pathway upon introduction of dsRNA
Next, we expanded our investigation to CCA cells lacking nc886. To activate the PKR/eIF2α pathway, we transfected a dsRNA mimic Poly(I:C) into M139 cells (nc886-, P-PKR+, P-eIF2α+). For comparison, we included two cell lines, MMNK1 (nc886+, P-PKR−, P-eIF2α-) and M214 (nc886+, P-PKR+, P-eIF2α+) in these experiments.
Poly(I:C) activated PKR in all the cell lines tested (). The intact PKR/eIF2α pathway was again confirmed in non-malignant MMNK1 cells in which Poly(I:C) inhibited global protein synthesis via P-eIF2α (). In contrast, P-PKR did not further phosphorylate eIF2α nor decrease global protein synthesis in M214 and M139 cell lines (). In M214 cells, Poly(I:C) treatment and nc886 knockdown yielded the same result (compare and ). So, P-PKR failed to phosphorylate eIF2α in these two cells.
P-PKR induced by dsRNA activated the NF-κB branch but not the eIF2α branch in CCA cells
This raised a question as to whether P-PKR also failed to activate its other downstream events. As eIF2α and NF-κB are two representative downstream branches in the PKR pathway, we examined the NF-κB pathway (). Poly(I:C) treatment activated the NF-κB pathway in all the three CCA cell lines tested, including M214 and M139 where P-PKR failed to phosphorylate eIF2α. This NF-κB activation was PKR-dependent, as it was abrogated by 2-aminopurine (2-AP), a PKR inhibitor (). Thus, M214 and M139 CCA cells selectively blocked the eIF2α branch, but not the NF-κB branch in the PKR pathway.
The NF-κB pathway in CCA cells
As the activation of NF-κB by P-PKR was operative in the presence of dsRNA, we questioned if the intrinsic NF-κB activity was elevated in CCA cells (P-PKR+) relative to MMNK1 cells (P-PKR−). First, we wanted to ensure that PKR activity was indeed higher in CCA cells than in MMNK1 cells by employing in vitro kinase assays. Consistent with our earlier P-PKR Western data (), M214 and M213 cells (P-PKR+) had a robust PKR kinase activity, compared to the very low activity of the MMNK1 cell line (P-PKR−) (). Accordingly, the NF-κB activity of these two CCA cells was higher than that of MMNK1, as assessed by luciferase assays measuring the NF-κB-responsive promoter activity () and by qRT-PCR measurement of NF-κB target mRNAs ().
NF-κB was elevated by P-PKR in CCA cells and played a pro-proliferative role therein
The dependence of NF-κB activity on P-PKR (thus PKR's kinase activity) was reinforced by knockdown experiments. In M156 and M214 (both nc886+), nc886 knockdown resulted in PKR activation in our earlier data (). Consistently, NF-κB was activated, as shown by qRT-PCR of NF-κB target genes (). This activation was significantly decreased by simultaneous knockdown of PKR by siRNA, demonstrating the requirement of PKR for the NF-κB activation by nc886 depletion.
Finally, we examined if the elevated PKR/NF-κB activity played a role in the proliferation of CCA cells. For this, we treated cells with caffeic acid phenethyl ester (CAPE), a chemical inhibitor for NF-κB, and measured cell proliferation. CAPE treatment suppressed proliferation of M214 and M213 (both P-PKR+) CCA cells (), demonstrating NF-κB's pro-proliferative role therein. A similar effect of CAPE on CCA growth has been reported previously (31
). In agreement, inhibition of PKR activity by 2-AP treatment also attenuated the proliferation of these CCA cells ().
The elevated P-eIF2α in CCA cells
The intrinsic expression of P-eIF2α in CCA cells was still intriguing but not addressed yet. An important obvious question was what caused eIF2α to be phosphorylated. The primary candidate was certainly P-PKR, an eIF2α kinase, because its expression was positively correlated with P-eIF2α (). However, this was questionable, because eIF2α was not phosphorylated by P-PKR in CCA cells upon nc886 knockdown or dsRNA treatment (-). To test this, we suppressed PKR in M214 and M213 cell lines (both P-PKR+, P-eIF2α+). siRNAs against PKR successfully decreased total PKR and P-PKR, but not P-eIF2α (). Thus, it is unlikely that P-eIF2α was induced by P-PKR in these cell lines.
The intrinsic expression of P-eIF2α, which was independent of P-PKR, was bypassed by eIF2B overexpression in CCA cells
Another candidate was MetAP2(p67), as it is known to bind eIF2 and protect eIF2α from phosphorylation by P-PKR [reviewed in (32
)]. We found its expression to be higher in CCA cells than in MMNK1 (). These data were against our scenario that eIF2α would be released from its protection and hyper-phosphorylated when MetAP2(p67) decreased in CCA. Our data suggested that P-eIF2α was induced in CCA cells by some factors other than P-PKR or MetAP2(p67).
An equally important question was how CCA cells could proliferate in the presence of P-eIF2α. Obviously, their protein translation normally occurred in spite of P-eIF2α's intrinsic expression (), as well as its induction by nc886 knockdown (M156 in ). P-eIF2α's mechanism for inhibition of protein translation is to bind to eIF2B with an excessively high affinity and consequently deplete productive eIF2B, an essential factor for protein translation (33
). Accordingly, overexpression of eIF2B has been shown to negate P-eIF2α's inhibitory effect on protein translation (34
). In addition, high expression of eIF2B in malignancies has been reported (18
). We measured eIF2B-ε, the catalytic subunit of the penta-peptide eIF2B complex (37
), and found it to be elevated in CCA cell lines relative to MMNK1 (). Thus it is likely that compensating high levels of eIF2B was one mechanism by which CCA cells overcame P-eIF2α.
nc886 and P-PKR expression in clinical samples
Thus far, we have shown the significance of nc886 in the PKR pathway in the cholangiocyte/CCA cell culture. To corroborate our in vitro findings, we measured nc886 in clinical specimens from CCA patients. Among 16 pairs of CCA tumors and adjacent normal tissues, nc886 decreased in tumor in eight cases, increased in four cases, and did not change significantly in four cases (). This pattern was very similar to that observed in the seven CCA cell lines (). Compared to MMNK1, nc886 decreased in four CCA cell lines and increased in one CCA cell line.
nc886/P-PKR expression in clinical specimens from CCA patients
Next, we measured P-PKR in another set of eight CCA samples with matched normal liver tissue samples by immunohistochemistry (IHC). P-PKR was strongly expressed in CCA but not at all in surrounding fibrous tissues (). In six of eight patients, the P-PKR expression in CCA was stronger than in matched bile ducts or liver tissues (). Thus, we have provided evidence here that the nc886/P-PKR expression in CCA patients was similarly altered to that observed in the CCA cell lines (). Although accumulation of more clinical data will be needed to assess nc886/P-PKR's in vivo significance more accurately, we suggest that the nc886/PKR tumor surveillance mechanism could operate in vivo.