The Malat1 long non-coding RNA is upregulated by signalling through the PERK axis of unfolded protein response during flavivirus infection

Flavivirus infection causes host cell death by initiation of an unfolded protein response (UPR). UPR is initiated following activation of three ER-membrane resident sensors, PERK, IRE1α and ATF6, which are otherwise kept inactive through association with the ER-chaperone GRP78. Activation precedes cellular and molecular changes that act to restore homeostasis but might eventually initiate apoptosis. These changes involve influencing function of multiple genes by either transcriptional or post-transcriptional or post-translational mechanisms. Transcriptional control includes expression of transcription factor cascades, which influence cognate gene expression. Malat1 is a long non-coding RNA which is over-expressed in many human oncogenic tissues and regulates cell cycle and survival. In this report, for the first time we show activation of Malat1 following infection by two flaviviruses, both of which activate the UPR in host cells. The temporal kinetics of expression was restricted to later time points. Further, Malat1 was also activated by pharmacological inducer of UPR, to a similar degree. Using drugs that specifically inhibit or activate the PERK or IRE1α sensors, we demonstrate that signalling through the PERK axis activates this expression, through a transcriptional mechanism. To our knowledge, this is the first report of an UPR pathway regulating the expression of an lncRNA.

sensors for binding to GRP78 molecules, thereby activating them. The activated sensors transduce the signal to different parts of the cell initiating the UPR. As part of UPR multiple transcription factors (TFs) are either activated or synthesized de novo, a few of them initiating a cascade of TF expression 6 . Additionally, the load of nascent proteins in the ER-lumen for folding is decreased through an inhibition of the ER-associated mRNA translation. This is achieved through phosphorylation of the translation initiation factor eIF2α by PERK and degradation of mRNAs that are associated with ER-membrane bound ribosomes by IRE1α 6,7 . In addition to inhibition of translation, activation of PERK also stimulates the transcriptional activity of NFE2L2 8,9 . The phosphorylation of eIF2α , attenuates rate of translation initiation on most mRNAs, but increases that for Atf4, which codes for the transcription factor (TF) ATF4 10 . One of the known target genes activated by ATF4, Chop or Ddit3, also codes for a TF with unique target genes 11 . However, in addition to the PERK axis, expression of Ddit3 is also influenced through the activity of other UPR axes 12 . The over-expression of DDIT3 in the course of UPR which is induced by JEV infection, has been demonstrated to be of particular relevance with respect to the consequential apoptotic death of the infected cells 3 . In consonance, these three well characterized TFs (ATF4, NFE2L2 and DDIT3), drive the expression of multiple genes.
Long non-coding RNAs (LncRNA) from varied gene loci are increasingly being reported to function as crucial regulators of gene expression 13 . The mode of regulation can be either transcriptional through alteration of chromatin or post-transcriptional through influencing the choice of alternative splice sites or functioning as a sponge for specific microRNA(s) [13][14][15][16] . Recent reports have also indicated specific lncRNAs to have roles in cell survival 17,18 . The Malat1 LncRNA has been reproducibly associated with aggressive carcinoma presenting a poor prognosis 19,20 . The physiological role of Malat1 has been shown to involve cell growth, cell migration and cell cycle 19,21 . However, mice knock-out for Malat1 did not exhibit any developmental aberrations 22 .
In this report we show Malat1 to be upregulated by JEV infection of mouse neuroblastoma cells Neuro2a, presumably through an induction of UPR. As proof of that, Malat1 was also upregulated by the pharmacological agent of UPR induction, thapsigargin or TG. Using different pharmacological drugs that either inhibit or activate specific UPR sensors, we present evidence that this upregulation is transcriptional and downstream of the signalling of the PERK axis of UPR.

Cell lines, virus infection and drug.
Neuro2a and mouse embryonic fibroblast (MEF) cells were maintained in DMEM supplemented with 10% foetal-calf serum at 37 °C and 5% CO 2 . Japanese encephalitis virus (Vellore strain) and West Nile virus were grown in Porcine kidney cells (PS) or Vero cells as described earlier 23 . For the purpose of infection, virus stocks were diluted in DMEM-2% FCS and incubated with cells for 1 hour. At the end of infection, the inocula were discarded and complete growth medium overlaid on infected cells.
For treatment with different drugs complete media supplemented with 1μ M of Thapsigargin (Sigma) either alone or supplemented with DMSO (appropriate concentration) or 150 nM of PERKi (Millipore) or 50μ M of STF083010 (Sigma) or 1μ g/ml of Actinomycin-D (Sigma), was overlaid on cells and incubated for the time indicated. For PERK activation cells were treated with complete media supplemented with 10μ M CCT020312 (Millipore) for the indicated time.
RNA isolation and real time PCR. Total RNA was extracted with Trizol reagent and purified using Qiagen RNeasy columns with in-column DNase digestion according to manufacturer's protocol. The RNA was reverse-transcribed using 50 ng of random hexamers and Improm-II reverse transcriptase (Promega) according to manufacturer's instructions. The cDNA was used for qPCR using 2x SYBR-green mix (ABI) in an ABI7500-Fast Real-time PCR machine. The primers used for qPCR are as follows: Immunoblot. Neuro2a cells treated with different drugs were washed with phosphate buffered saline and the total protein extracted using Cellytic lysis reagent (Sigma) supplemented with complete protease inhibitor cocktail (Roche). The protein concentration in all extracts were estimated using Bradford reagent (Biorad) and equal quantity of protein resolved using SDS-8%PAGE followed by immunoblotting. The immunoblot for total PERK was performed using anti-PERK antibody (C33E10; Cell Signalling) and that for phosphorylated-PERK using anti-phospho PERK antibody (16F8; Cell Signalling) according to manufacturer's instructions. The immunoblots were visualized using western blotting luminol reagent (Santa Cruz).

Results
Malat1 is overexpressed upon JEV and WNV infection. A study of the alterations in transcriptome in Neuro2a cells upon JEV infection showed a moderate but statistically significant (P < 0.01) over-expression of the Malat1 long non-coding RNA, which was validated by real-time PCR analysis (data not shown). Further, a study of the temporal kinetics of this overexpression showed the activation to ensue between 12 and 24 h post-infection (p.i.) (Fig. 1, panel A). In addition to Neuro2a infected by JEV, mouse embryonic fibroblasts (MEFs) infected with either JEV or West Nile virus (WNV), a close relative of JEV, also showed statistically significant augmentation of Malat1 transcript level ( Fig. 1, panel B). The host cell responses that have been characterized following flavivirus infection include innate anti-viral response and unfolded protein response. In general innate immune anti-viral responses are early, while gene expression changes that follow UPR induction in flavivirus infected cells, has been shown to be induced about 12 h p.i 4 . Since the upregulation of Malat1 was observed between 12-24 h p.i., we presumed this gene activation to be a part of the UPR, which is activated in Neuro2a cells following JEV infection. In accordance with such a supposition, we observed a similar degree of upregulation of the Malat1 transcript in cells that were treated with thapsigargin (TG), a pharmacological inducer of UPR. (Fig. 1, panel C). In order to assess the generality of this observation, the effect of TG-treatment in regulating Malat1 level was tested on three other cell lines, two of them being of mouse origin (Mouse embryonic fibroblast cells or MEFs and the mouse microglial cell line BV2) and one of human origin (Human embryonic kidney cells or HEKs). Interestingly, although TG-treatment showed a similar effect on the Malat1 level in MEFs, it did not induce Malat1 level in either BV2 or HEKs (Fig. 1, panel D and data not shown). Expectedly, JEV infection of BV2 also did not induce augmentation of Malat1 level (data not shown).
Malat1 is over-expressed as part of the UPR downstream of the PERK sensor. The gene expression programme characteristic of UPR is driven by activation of three ER-resident sensors. The activation of PERK induces autophosphorylation followed by phosphorylation of eIF2α , which in turn induces a global inhibition on protein synthesis. This autophosphorylation of PERK, and thereby the subsequent steps, can be inhibited by PERKi (Fig. 2, panel A) 24 . A similar autophosphorylation is observed subsequent to stimulation of the IRE1α sensor 25 . Phosphorylation activates an incipient RNase activity in IRE1α which performs a unique cytoplasmic splicing of the transcript coding for XBP1U (Xbp1u) to produce the transcript that can code for the TF XBP1S (Xbp1s) 25 . The RNase activity of IRE1α can be inhibited by the specific inhibitor STF083010 (Fig. 2, panel A) 26 . Co-treatment of Neuro2a cells with TG and different concentration of PERKi showed a total inhibition of the TG-mediated PERK phosphorylation in Neuro2a by 150 nM of the inhibitor (Fig. 2, panel B, compare lanes 1, 2 and 4). Previously, we have demonstrated inhibition of IRE1α RNase activity upon addition of STF083010 to Neuro2a cells 27 . A pharmacological inhibition, of either of these pathways in TG-treated Neuro2a cells using the respective drugs, showed the signalling initiated by PERK but not that by IRE1α to be responsible for Malat1 upregulation (Fig. 2, panel C). An inhibition of PERK activation by PERKi would suggest reduced synthesis of ATF4, which would be evident through a reduction in the transcription of its cognate target genes. As further proof of a block in signalling downstream of activated PERK upon addition of PERKi, we observed a reversal of TG-mediated over-expression of the ATF4 target gene Ppp1r15a (Fig. 2, panel D) 28 . It is possible to pharmacologically activate only the PERK sensor, without sensitizing the other, by treatment of cells with the drug CCT020312 (CCT) 29 . As further confirmation of the involvement of PERK signalling, activation of only the PERK axis by CCT showed increase in Malat1 level (Fig. 2, panel E). As a proof of activation of the PERK axis we observed upregulation of the ATF4 target genes Ppp1r15a (Fig. 2, panel E) 28 .
The increase in Malat1 level is transcriptional. As part of an UPR, alterations in the expression of genes coding for both proteins and non-coding regulatory RNAs like microRNAs, is observed 30 . Further, Malat1 has been shown to be amenable to negative regulation by three different miRNAs 31,32 . Therefore, to delineate the observed upregulation as either transcriptional or post-transcriptional, Malat1 levels were compared between cells treated with either TG alone or TG in presence of Actinomycin-D (Act-D), a known inhibitor of RNA polymerase activity. The results showed that the upregulation of Malat1 by TG-treatment can be reversed by Actinomycin-D (Fig. 3). As control, we observed similar reversal for transcripts coding DDIT3 and DNAJC3, which are respectively activated transcriptionally by the UPR-related transcription factors ATF4 and XBP1 (Fig. 3) 33,34 . Further, as noted earlier, treatment with TG in the presence of Act-D, led to an even higher repression of St3gal5 transcript, which is degraded by the IRE1α as part of RIDD pathway of gene regulation (Fig. 3) 35 . A treatment of Neuro2a cells Act-D alone repressed Malat1 level to those observed upon simultaneous addition of Act-D and TG, indicating a continuous transcription of the gene under basal conditions, which is augmented during UPR in a PERK dependent manner.

Discussions
The gene expression changes in a virus-infected host cells can potentially have an antiviral or a proviral effect. The error-prone genome replication of RNA viruses provides molecular variants that can exploit a favourable change and occupy the respective niche, thus driving viral evolution 36 . We have previously shown how UPR-specific gene regulatory pathways can have a beneficial effect on JEV replication 27 . The infection induced UPR has also been implicated as the gene expression program that drives cell death during JEV infection, by signalling downstream of the PERK axis 3 . In this report we show Malat1, an   -6), was resolved in SDS-PAGE and immunoblotted using either anti-PERK or anti-phospho-PERK antibody. The numbers on the left indicate the migration of relative protein molecular weight marker bands. (C) Total RNA was isolated from Neuro2a cells treated with either DMSO (white bar) or TG (grey bar) or TG supplemented with PERKi (hashed bar) or TG supplemented with STF083010 (black bar), at 6 h after addition of drugs. The total RNA was reverse-transcribed and used for qPCR estimation of Malat1 transcript level, normalized to that of Gapdh mRNA. The relative transcript level in DMSO-treated cells was taken arbitrarily as 1 and that in others represented as fold-change over it. (D) Total RNA was isolated from Neuro2a cells treated with either DMSO (white bar) or TG (grey bar) or TG supplemented with PERKi (black bar), at 6 h after addition of drugs. The total RNA was reversetranscribed and used for qPCR estimation of Ppp1r15a transcript level, normalized to that of Gapdh mRNA. The relative transcript level in DMSO-treated cells was taken arbitrarily as 1 and that in others represented as foldchange over it. (E) Total RNA was isolated from Neuro2a cells treated with either DMSO or CCT020312 at 6 h after addition of drugs. The total RNA was reverse-transcribed and used for qPCR of the respective gene transcripts, normalized to that of Gapdh mRNA. The transcript level in DMSO-treated cells was arbitrarily taken as 1 (white bar) and that from CCT020312-treated cells (grey bar) represented as fold-change over it. All experiments were done in triplicates and the error bars represent the standard deviation (** signifies p < 0.01).
lncRNA previously established to play a critical role in metastasis of cancer cells, to be upregulated in JEV-infected Neuro2a cells. Further, Malat1 was activated by pharmacological induction of UPR, through signalling downstream of the PERK sensor. In a manner similar to many other genes that are activated by this sensor, the upregulation of Malat1 was transcriptional. To our knowledge, this is the first report that shows the Malat1 gene to be upregulated following activation of a specific ER-sensor during ER-stress.
Malat1 upregulation has been associated with metastasis of cancer cells, in which it has been shown to promote migration/invasion, an effect which can be reversed by RNAi-mediated silencing of the transcript 32,37 . Recent reports have implicated hypoxia as a trigger for activating Malat1 expression, a condition which has also been reported to activate PERK signalling 38,39 . In separate reports the TGFβ pathway, the c-fos and Sp1 transcription factor have been shown to promote Malat1 expression 20,40,41 . Also, endogenous post-transcriptional regulation of Malat1 transcript by miRNAs has been demonstrated 31 . An exogenous depletion of Malat1 was demonstrated to induce cell cycle arrest in G2/M phase with concomitant augmentation of apoptosis 20,37 .
The functions of Malat1 have been implicated to be principally in regulating gene expression in addition to controlling alternative splicing, in a manner which is dependent on localization of the transcript to nuclear speckles, although it is not required for the integrity of nuclear speckles per se 15,[42][43][44] . An association of this lncRNA with pre-mRNAs at active transcription loci, which leads to modulation in the expression of cell cycle regulating transcription factor genes, has been demonstrated 14,21,45 . In a manner that is relevant to understanding any plausible significance of Malat1 over-expression in virus infected cells, this lncRNA has been shown to not have any significantly influential role in expression of interferon induced genes 43 .
At the cellular level Malat1 transcript promotes inflammatory response, possibly through activation of the PI3/Akt and ERK/MAPK pathways [46][47][48] . At the molecular level the Malat1 transcript has been shown to function as a competing endogenous RNA by acting as a miRNA-sponge for miR-133 49 . Previously, we performed differential expression analysis for microRNAs in JEV-infected Neuro2a cells versus mock-infected controls. The results showed no detectable expression of miR-133 in this cell line, thus negating the possibility of such a regulatory role for Malat1 (data not shown).
Reports from multiple groups suggest ER-stress signalling to modulate virus infection cycle. In fact, ER-stress has been shown to enhance the Interferon pathway of innate anti-viral immune response 50 . Activation of lncRNA genes in different model system of host-pathogen interactions has been demonstrated earlier. Interestingly, Saha and co-workers showed upregulation of certain lncRNAs in mouse brain following JEV infection, one of which was Neat1 (referred to as VINC in the article) 51 . Enhanced expression of Malat1 has also been reported from clinical samples infected with human papilloma virus (HPV) 52 . In addition to JEV infection, HIV infected cells also showed over-expression of Neat1, another lncRNA which is expressed from a gene loci very close to that of Malat1 53 . Further, Neat1 expression has been shown to have an anti-viral role in suppressing HIV replication, although no role in JEV infection is known 53 . Since both Neat1 and Malat1 have been shown to localize to active transcribing loci in the chromatin, this might suggest a potential anti-viral role for Malat1 as well 14 . However, the implication of Malat1 over-expression for UPR in general and virus-induced UPR in particular is still not clear yet, and would need further investigation. Total RNA was isolated from Neuro2a cells treated with DMSO or TG or TG supplemented with Actinomycin-D or Actinomycin-D alone, at 6 h after the addition of the drugs. The RNA was reverse-transcribed and used for qPCR estimation of transcripts from respective genes normalized to that of Gapdh. The relative transcript level in DMSO-treated cells (white bar) was arbitrarily taken as 1 and that in cells treated with TG (grey bar) or TG supplemented with Actinomycin-D (hashed bar) or Actinomycin-D alone (black bar) represented as fold-change over it. All experiments were done in triplicates. The error bars represent the standard deviation.