Complex I Binding by a Virally Encoded RNA Regulates Mitochondria-Induced Cell Death

See allHide authors and affiliations

Science  01 Jun 2007:
Vol. 316, Issue 5829, pp. 1345-1348
DOI: 10.1126/science.1142984


Human cytomegalovirus infection perturbs multiple cellular processes that could promote the release of proapoptotic stimuli. Consequently, it encodes mechanisms to prevent cell death during infection. Using rotenone, a potent inhibitor of the mitochondrial enzyme complex I (reduced nicotinamide adenine dinucleotide–ubiquinone oxido-reductase), we found that human cytomegalovirus infection protected cells from rotenone-induced apoptosis, a protection mediated by a 2.7-kilobase virally encoded RNA (β2.7). During infection, β2.7 RNA interacted with complex I and prevented the relocalization of the essential subunit genes associated with retinoid/interferon–induced mortality–19, in response to apoptotic stimuli. This interaction, which is important for stabilizing the mitochondrial membrane potential, resulted in continued adenosine triphosphate production, which is critical for the successful completion of the virus' life cycle. Complex I targeting by a viral RNA represents a refined strategy to modulate the metabolic viability of the infected host cell.

During primary infection or reactivation of human cytomegalovirus (HCMV), especially in the immunocompromised, the virus is able to replicate in a number of cell types, often resulting in life-threatening disease (1). HCMV exhibits a relatively protracted life cycle (upwards of 5 days) and at early times of infection (12 to 24 hours) encodes a highly abundant 2.7-kb RNA transcript (β2.7), accounting for >20% of total viral gene transcription (2, 3) of unknown function. The RNA may be associated with mitochondria (4), and no protein product of this RNA has ever been detected in infected cells (3), suggesting that it functions as a noncoding RNA (5).

We investigated the possibility that β2.7 could function as a noncoding RNA. A Northwestern screen of a human cDNA library with a β2.7 probe identified potential cellular interaction partners for the β2.7 RNA molecule. One of these proteins was a subunit of the mitochondrial enzyme complex I (reduced nicotinamide adenine dinucleotide–ubiquinone oxido-reductase). Defective complex I activity has been implicated in numerous mitochondrial and genetic diseases, including Leigh's syndrome, Leber's hereditary optic neuropathy, and mitochondrial encephalopathy (6); and inhibition of complex I activity by reactive O or N species or by the direct binding of environmental toxins ultimately results in apoptosis (7, 8).

We first tested whether HCMV infection [multiplicity of infection (MOI) = 5] and, specifically, β2.7 expression prevented cell death in neuronal U373 cells subjected to mitochondrial stress by treatment with rotenone—a highly effective complex I inhibitor (9, 10). As expected, the addition of rotenone promoted substantial cell death in U373 cells (70% in Fig. 1). We then compared the effect of preinfection of cells with the Toledo strain of HCMV and a recombinant Toledo virus, in which the β2.7 gene had been deleted (Δβ2.7Tol) (5). Toledo-infected U373 cultures showed profoundly (P < 0.001) reduced levels of apoptosis (3%), in contrast to Δβ2.7Tol-infected cells (70% in Fig. 1A). Furthermore, the protective effect of β2.7 could be restored with a revertant virus (Fig. 1A) and by transfection of a β2.7 expression vector into U373 cells (Fig. 1B). Thus, HCMV-mediated protection of cells from rotenone-induced apoptosis correlated with expression of the viral β2.7 gene. Although the consensus is that the β2.7 transcript does not encode a protein product (5), we also analyzed a clinical isolate of HCMV (HCMV-3157), which would produce a heavily truncated nonsense protein if β2.7 were translated (5). HCMV-3157 was as efficient as Toledo at protecting cells from rotenone-induced death (fig. S1A). The protective effect of β2.7 was observed only in Toledo-infected cells (fig. S1B), whereas Δβ2.7Tol-infected cells routinely stained for both viral gene expression and terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) (fig. S1B). The proportion of immediate-early (IE)–positivity in the Δβ2.7Tol cells undergoing apoptosis was around 70% (fig. S1C). Given that only 45% of the whole population was infected, this finding implied that viral infection, in the absence of β2.7, actually rendered cells more sensitive to rotenone-induced apoptosis.

Fig. 1.

HCMV protects cells from rotenone-induced apoptosis. (A) Percentage of TUNEL-positive cells in mock- (1 and 2), Toledo- (3 and 4), Δβ2.7Tol- (5 and 6), and Δβ2.7Tol-Revertant (Δβ2.7Tol-Rev)–infected (7 and 8) U373 cells incubated with acetone (1, 3, 5, and 7) or rotenone (ROT) (2, 4, 6, and 8). n = 3 independent experiments. Asterisks denote P < 0.01 with Student's t test. Error bars indicate 1 SEM. (B) Percentage of TUNEL-positive pREP10- (Con) (1) and pREP10-β 2.7–transfected (2) cells incubated with rotenone. from three independent analyses. Reverse transcription (RT)–PCR for β2.7 expression in pREP10- (lane 1) and pREP10-β2.7–transfected (lane 2) cells is shown.

One role of the HCMV β2.7 transcript may be to mediate protection of the cell from apoptotic pathways activated by metabolic stress of complex I. Genes associated with retinoid/interferon–induced mortality (GRIM)–19 is a subunit of complex I that is essential for its assembly and function (11). HCMV infection up-regulates steady-state mRNA levels of some subunits of mitochondrial complexes I to V (12), but virus infection has no impact on GRIM-19 protein expression, up to 120 hours post infection (hpi) in U373 cells (fig. S2, A and B). We did, however, detect changes in GRIM-19 localization in response to rotenone and virus infection. In U373 cells, GRIM-19 expression appeared to be diffuse throughout the cytoplasm of the cell (Fig. 2A), and it became relocalized into discrete perinuclear clumps after the addition of rotenone to uninfected U373 cells. However, preinfection with Toledo prevented this relocalization (Fig. 2A). GRIM-19 is known to localize to the nucleus under certain conditions (13), and it also interacts with the signal transducer and activator of transcription (STAT)–3 protein to prevent the nuclear import of STAT-3 in a perinuclear location (14). Although the GRIM-19 in Toledo-infected U373 remained predominantly cytoplasmic within the cell, it did exhibit a more punctuate pattern within the cytoplasm (Fig. 2A), perhaps suggesting more association with mitochondria. However, HCMV itself has profound effects on mitochondrial shape and localization after infection (15), and this may also partially account for the difference in GRIM-19 staining. In contrast, in Δβ2.7Tol-infected cells, the rotenone-induced perinuclear relocalization of the GRIM-19 protein was still observed (Fig. 2A), identical to that observed with uninfected cells treated with rotenone (d in Fig. 2A). Thus, rotenone-induced mitochondrial stress promotes the relocalization of GRIM-19 within the cell, which can be abrogated by expression of the HCMV β2.7 transcript.

Fig. 2.

HCMV interacts with and prevents the rotenone-induced relocalization of, GRIM-19 in virally infected cells. (A) GRIM-19 localization (red) in mock- (a and d), Toledo- (b and e), and Δβ2.7Tol-infected (c and f) U373 cells incubated with solvent control (Con) (a, c, and e) or rotenone (b, d, and f) 24 hpi. (B to D) Immunoprecipitation of control immunoglobulin G (lane 2), GRIM-19 (G-19) (lane 3), native complex I (C-1) (lane 4), and native complex V (C-V) (lane 5) from Toledo- (B and C) or Δβ2.7Tol-infected (D) cells, and RT-PCR for β2.7 [(B) and (D), lanes 1 to 5], IE72 (C), or a β2.7 PCR with no prior RT [(B), lanes 6 to 10]. Inputs are shown in lane 1. (E) IPs on Toledo-infected cells with anti-sense oligonucleotides to β2.7 (lane 3) or IE72 (lane 4) or with no oligonucleotide (N) (lane 2). Silver stain of 1% input (INP) (lane 1), the immunoprecipitated proteins (lanes 2 to 4), and known protein loading controls (5 to 500 ng, lanes 5 to 7) is shown. (F) GRIM-19 expression in 10% input (lane 1) and the β2.7 (lane 2) and IE72 (lane 3) RNA-IP samples.

We next tested whether this effect on GRIM-19 relocalization was due to a physical interaction between the GRIM-19 protein and β2.7 RNA during infection. β2.7 RNA was specifically immunoprecipitated from infected cells with An anitbody to GRIM-19 (Fig. 2B). We also observed an interaction with native complex I, which is found only on the inner mitochondrial membrane, suggesting that the interaction between β2.7 RNA and GRIM-19 targets complex I in mitochondria. No immunoprecipitation of a similarly abundant viral RNA (IE72 in Fig. 2, B and C) with GRIM-19, complex I, or complex V (Fig. 2C) was observed. As expected, analyses using the Δβ2.7Tol virus showed no β2.7-specific polymerase chain reaction (PCR) band (Fig. 2D).

In a reciprocal analysis, we captured β2.7 RNA with the use of biotin-labeled oligonucleotide probes (Fig. 2E). Immunoprecipitations (IPs) of the β2.7-captured complexes did contain GRIM-19 (Fig. 2F). Thus, β2.7 RNA specifically interacts with GRIM-19, but with few other proteins in vivo.

The observed physical interaction with GRIM-19 and complex I was investigated further. Active complex I supports the formation of an electrochemical gradient (ΔψM) across the inner mitochondrial membrane, which is imperative for the efficient production of adenosine triphosphate (ATP) (16). Expression of β2.7 RNA during infection or transfection protected ΔψM stability from rotenone (fig. S3, A and B), suggesting that the β2.7 RNA interaction with complex I could affect mitochondrial energy production under oxidative stress after infection. Previous data have shown that, unlike in herpes simplex virus (17), ATP levels in HCMV-infected cells are maintained at 24 hpi (18), a requirement probably attributable to HCMVs' comparatively protracted growth phase (19). We therefore analyzed the role of β2.7 in ATP production.

Rotenone substantially reduced ATP production in U373 cells (75% reduction, 1 and 2 in Fig. 3A). However, in Toledo-infected cells, only a 1.2-fold reduction was observed (84 to 67% in Fig. 3A), suggesting that HCMV protected ATP production in infected cells. In Δβ2.7Tol-infected cells, rotenone treatment resulted in a 2.5-fold depletion of intracellular ATP (77 to 33% in Fig. 3A). At 6 hpi, before β2.7 RNA expression, no protection from rotenone-induced ATP depletion occurred (fig. S4). Using a second strain of HCMV AD169 that encodes two copies of the β2.7 gene, we eliminated the possibility that green fluorescent protein (GFP) expression from the β2.7 gene locus would have a phenotypic effect, because a recombinant AD169 virus still expressing GFP and one functional copy of the β2.7 gene maintained ATP levels (Fig. 3A). Thus, the β2.7 transcript is important for maintaining ATP production.

Fig. 3.

β2.7 expression maintains ATP production in infected cells. (A) Percentage of ATP content of mock- (1 and 2), Δβ2.7Tol- (3 and 4), Toledo- (Tol) (5 and 6), AD169-, and AD169-GFP–infected cells with (2,4, 6,8, and 10) or without (1, 3, 5, 7, and 9) rotenone for 2 hours was determined at 48 hpi, as compared to mock (1). (B) Percentage of ATP in mock-(1), Δβ2.7Tol- (2), Toledo- (3), AD169- (4), or AD169-GFP–infected cells (5) 5 dpi. n = 3 independent experiments. Asterisks denote P < 0.01 with Student's t test. Error bars indicate 1 SEM.

We hypothesized that the impact of β2.7 expression on ATP production may be more profound at later times of infection. Analysis in the absence of rotenone showed that ATP levels in Δβ2.7Tol dropped significantly (P < 0.01) at 5 days post infection (dpi), in direct contrast with Toledo-infected cells. Confirmation that the difference was not an artifact of GFP production was performed with the AD169-GFP–infected cells, in which the levels of ATP were comparable to those of the parent virus AD169 (Fig. 3B).

Deletion of the viral β2.7 gene from HCMV has no significant effects on growth kinetics in fibroblasts (5) (Fig. 4A). At first, this appears to be contradictory; however, fibroblasts are particularly resistant to the induction of apoptosis by rotenone and oxidative stress (20), as compared with neuronal cells (21). However, a Δβ2.7Tol growth defect was observed when compared with the Toledo virus (Fig. 4B), which was more profound in the presence of rotenone (Fig. 4B), as is entirely consistent with the β2.7 transcript supporting virus production in times of metabolic stress. Tissue culture was performed in glucose-enriched media, and fibroblasts, particularly in times of diminished ATP production from the electron transport chain (ETC), use this additional glucose to generate ATP via alternate pathways (22). We observed that glucose-depleted media impaired the growth of the Δβ2.7Tol virus in U373 and fibroblast cells, as compared with Toledo (Fig. 4C). This finding correlated with a drop in ATP production in Δβ2.7Tol-infected fibroblasts (fig. S5A) and the relocalization of GRIM-19 in Δβ2.7Tol-infected, but not Toledo-infected, cells (fig. S5B). Because cells in vivo are not exposed to such artificially high levels of glucose and are more reliant on the ETC for ATP production, the effects of β2.7 expression in vivo on metabolism are probably more overt than they are during tissue culture.

Fig. 4.

Toledo, but not Δβ2.7Tol, grows normally in metabolically stressed cells. (A) Growth of Toledo (triangle), AD169 (diamond), AD169-GFP (square), or Δβ2.7Tol (cross) in human fibroblasts (HFs). PFU, plaque-forming units. (B) Growth of Toledo (diamond and triangle) or Δβ2.7Tol (square and cross) in U373 cells. At 24 hpi, Toledo-infected (triangle) and Δβ2.7Tol-infected (cross) U373 cells were incubated with rotenone (asterisk). (C) Growth of Toledo (square and cross) or Δβ2.7Tol (diamond or triangle) in U373 and HF cells in glucose-depleted media.

Metabolic dysfunction, the breakdown of mitochondrial integrity, and the release of proapoptotic stimuli are hallmarks of mitochondria-induced apoptosis (23). HIV-1 targets complex I for degradation promoting apoptosis (24), presumably by promoting the formation of reactive oxidative species. Because of the pivotal nature of mitochondria in cell death, it is not surprising that HCMV targets mitochondria function and makes a concerted effort to subvert the apoptotic response including UL36 (caspase inhibitor), UL37x1 (a B cell lymphoma 2 homolog with profound anti-apoptotic activity) (25, 26), and UL38, which protects infected cells from endoplasmic reticulum stress (27). UL37x1 is also known to promote mitochondrial membrane stability (26) and is predominantly active at late times of infection (>48 hpi) (28). In contrast, the β2.7 gene RNA is expressed much earlier during infection. Consequently, it is likely that HCMV has evolved multiple functions in order to hijack the mitochondria in the cell to enable continued energy production, as well as protection from cell death.

An intriguing aspect of this study is that an RNA molecule is used directly to exert these effects. Although at first appearing unusual, this may represent a highly refined viral strategy. First, because of the sheer numbers of mitochondria within the cell, the expression of a highly abundant RNA allows the virus to saturate these organelles effectively. Second, by disposing of the need to translate the superabundant β2.7 RNA, the virus can achieve this effect more quickly throughout the course of infection.

Supporting Online Material

Materials and Methods

Figs. S1 to S5


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article