These results reveal a novel mechanism for transcriptional regulation of the PURA gene that allows separation of promoter controls for cell growth signals and signals for response to viral infection. PURA transcription is initiated at three separate start sites that generate distinct transcripts represented in EST databases. Our PCR amplification of transcripts from tissues and cell lines confirms initiation at each of these sites and reveals that each of the three transcripts is differently spliced. Promoter usage and alternative splicing differs among 8 different human tissues examined. Homologies between these promoters in human and mouse and the presence of three mouse EST transcripts similar to the human ones, render it likely that PURA transcriptional control is similar in both human and mouse. Human PURA TSS I is furthest upstream of the translational start site. Because its promoter is very homologous to a previously characterized mouse promoter , we have not further studied it here. In the mouse, this TSS I promoter is notably subject to feedback regulation by the Purα protein . Human TSS II is adjacent to an intriguing cluster of elements potentially involved in an innate immune response to viral infection. We have thus focused here on characterization of promoters II and III. The results obtained are surprising and raise important questions potentially relevant to any gene regulated by multiple promoters. We address some of those questions here.
Clearly, higher eukaryotes have evolved a transcriptional control mechanism, conserved between human and mouse, whereby three different promoters generate transcripts specifying the Purα protein. What evolutionary advantage could that confer? One advantage could be that the three transcripts are utilized differently to express Purα protein. This type of regulation could be similar to that observed in polyomaviruses SV40 and JCV, in which the gene encoding large T-antigen is under control of multiple promoters. In those cases the different transcripts are translated at different rates [31–33]. Another potential advantage could be that the different PURA RNAs have specific functional properties. In the rapidly developing field of regulatory RNA, non-coding RNA, often derived from introns in 5'-UTRs, plays an important regulatory role in mRNA translation. It is conceivable that the primary aim of the PURA cellular transcriptional manipulation is to generate different non-coding RNA species.
Several viruses are reported to co-opt functions of Purα [12, 34]. Is the downregulation of Purα upon MCMV infection a cellular innate immune response, or is it an aspect of viral co-opting of a cellular process? In known instances where Purα function is co-opted by viruses, e.g., upon infection with HIV-1 or JC virus [1, 35], viral proteins bind to and alter the function of Purα. That may not be the case regarding certain aspects of the cellular PURA response to MCMV. The most dramatic decrease in PURA expression occurs at a time when a minimal number of MCMV proteins (immediate early proteins) are synthesized. It remains to be determined whether later aspects of the cellular response to MCMV involve co-opting of Purα functions.
The NIH 3T3 cellular response to MCMV infection involves a very rapid loss of PURA mRNA and Purα protein. What is the advantage of such a rapid response, and how is it achieved? Purα is a highly conserved protein that clearly plays a vital role in cell survival . Redundancy of Pur family members is further evidence for the essential nature of this family [19, 36]. The protective role of Purα against oncogenic transformation is well documented [13, 15, 17, 37], and it involves rapid changes in intracellular levels during the cell cycle [13, 15, 37]. Therefore, in order to either modulate or co-opt Purα function, rapid changes in the protein must ensue. It should be noted that the protein Rb is rapidly degraded in response to CMV infection [38, 39]. Rb is a well-known binding partner of Purα [13, 34, 37], and the two act together to regulate the cell cycle. It is thus conceivable that if Rb is subject to proteolysis, its binding partner Purα could be exposed to a similar process of degradation.
The rapid and nearly complete loss of Purα from 3T3 cells within one hour of MCMV infection is intriguing. MCMV could initiate signaling pathways shortly after binding to receptors on the cellular membrane. The integrin receptors are a known point of attachment for HCMV . The pathways may work through proteins, which are constitutively present in the cytoplasm and are integral to the innate immune response. These include proteins such as IRF-3, which becomes phosphorylated and thus activated upon virion binding to the cell . We have shown that IRF-3 binds the PURA promoter and down-regulates PURA transcription. Regulatory tegument proteins which are transported into the cytoplasm on infection may also be involved . The rapid decrease in levels of both PURA mRNA and Purα protein would most likely involve the degradation of both. To reach low levels within one hour, pre-existing PURA RNA and protein would need to be destroyed. A transcriptional response to interferon may not be involved. These findings are consistent with microarray data indicating that HCMV infection of monocytes led to a decrease in PURA mRNA at immediate early times during infection .
The significance in the decrease of Purα should be considered while keeping in mind what is known about Purα levels during the cell cycle. Purα levels fluctuate and reach their lowest level at the onset of S-phase . Elevated levels of Purα in NIH 3T3 cells delay entry into and progression through S-phase [15, 42]. It has been reported that MCMV infection alters the cell cycle causing rapid progression toward the G1/S boundary [43, 44]. Thus CMV may have evolved a mechanism to decrease Purα levels in order to further a viral propensity to replicate.
There are multiple processes involved in the PURA response to MCMV infection. Following the initial degradation of both Purα and PURA mRNA, there is a subsequent rapid  partial recovery of Purα intracellular levels. This could involve increased translation of existing PURA mRNA because at this time levels of such mRNA are still decreasing (Figure 8). It is notable that levels of Purα protein increase modestly from 3 hrs through 9 hrs post infection, although levels of its mRNA are decreasing until at least 5 hrs. It is during this time that an interferon response to viral mRNA would be expected. Levels of Pur mRNA and protein never recover to mock, i.e., uninfected, levels during this time. Thus, the decrease in overall levels of Purα in response to MCMV infection is consistent with our data indicating that IRF proteins repress PURA gene transcription. Experiments beyond the scope of this paper will be necessary to fully comprehend the multiple components of PURA response to viral infection.
The three PURA transcriptional sites identified here as TSS I, II, and III are approximately 6,404 bp, 1,249 bp, and 80 bp upstream from the translational start codon, respectively. Transcripts from TSSs I and II are characterized by the removal of introns of distinctive lengths: 5,939 nt from human TSS I transcripts and 830 nt from TSS II transcripts. No intron is removed from TSS III transcripts or from the PURA coding sequence. The question posed is whether these three sites are regulated by different transcriptional elements and with factors associated with different cellular processes. The role of the Group II TSS is very intriguing. An analysis of transcription from this site in 8 human tissues revealed that it contributes to the total PURA transcript differently in these tissues. Surprisingly, the intron is frequently not removed from the transcript, and the amount of intron splicing varies with the different tissues. Only in brain tissue is the removal of the intron the usual means of processing. Intron processing also varies between normal and cancerous tissue as seen when comparing lung total RNA to RNA from lung adenocarinoma (Figure 3). While intron RNA is readily detectable in both tissues, adenocarcinoma has a distinctly greater concentration of spliced transcript.
An inspection of potential regulatory elements lying close to TSS II reveals clusters of elements for binding heat shock factor I and factors associated with viral infection, such as interferon, IRF-1, NF-kappa B and the element ISRE which binds the IRF transcription factors. Considering that Purα is required in many processes involving ss-RNA or DNA, the presence of these binding sites suggests that the cell may actually optimize survival during viral infection by regulating the availability of Purα. The ability of the IRFs 3, 5, and 7 to repress transcription from pGL3-hPURAPr to as low as 27% of control, supports this hypothesis although these IRFs did not exclusively act near TSS II. IRF-9 appears to act primarily at the TSS II promoter (Figure 6). Using ChIP analysis, we demonstrate that IRF-3 binds a specific DNA site within the TSS II promoter. The ability of IRF-3, usually a positive transcriptional regulator, to down-regulate gene expression has previously been reported by Grandvaux et al. .
Analysis of the total RNA recovered from MCMV infected 3T3 cells demonstrates that PURA transcription is altered in response to viral infection. Strikingly, there is a significant decrease in the processing of TSS II transcript, which results in a large increase in the amount of transcript from which the intron has not been removed. The binding of IRF-3 and the large number of potential binding elements near TSS II suggest that this promoter is regulated by the interferon directed innate immune response to viral infection. The mechanism of the altered splicing in this case is unknown, but the overall effect is similar to that reported for splicing of various brain transcripts from genes with alternate promoters . Detailing the mechanism of promoter-specific altered mRNA splicing, beyond the scope of this study, will be an important future research subject.
The analyses of transcription from PURA TSS II in the human tissue panel and in cells transfected in culture showed that differing cellular environments result in variable amounts of transcription. Moreover, different portions of the promoter sequence can affect transcription differently in various cell types. This is seen when comparing transcription from pGL3-hPURAPr Δ(-1514 -1202) in Figure 4 where this deletion mutant gives a significant reduction in transcription relative to control vector, whereas in Figure 6, pGL3-hPURAPr Δ(-1514 -1202) yields an increase, not statistically significant, in transcription. It is notable that Figure 4 was done with small-cell lung carcinoma cell lines, each of which highly overexpresses c-MYC , and that Figure 6 used HEK 293T cells, which express SV40 large T-antigen (American Type Culture Collection). Each of those proteins are responsible for multiple cellular changes that can have different transactivational effects, including epigenetic changes in DNA methylation or histone modifications that can directly affect transcription of PURA and the deletion mutants . These various epigenetic changes affect chromatin conformation causing sequence to vary in its availability to transactivational factors. While elements typically bind factors to facilitate transcription, a different placement of elements could bind the same factors in an arrangement that results in steric hindrance and the suppression of transcription. In this way the IRFs might be used to deny infecting virions cellular proteins that are essential for their replication.