Viruses reproduce by entering host cells and converting them into factories for making new viral particles. This process involves hijacking of cellular gene expression machinery to promote replication of viral components and suppress the host cell’s defense mechanisms. To date, viral replication and its effects on host cell transcription in vitro have largely been studied in bulk populations using high infectious doses. While these practices are experimentally convenient, they prevent assessment of variability in host cell responses and may not be representative of naturally occurring infections, in which a handful of viral particles is sufficient to initiate infection.
The Bloom Laboratory in the Basic Sciences Division strives to capture a more accurate picture of viral infections by using state-of-the-art sequencing methods. “To understand—and potentially manipulate—viral disease, we must turn to approaches that allow us to see all of the activities viruses are engaging in, at the level of individual viruses infecting individual cells,” says Dr. Alistair Russell, a post-doctoral fellow in the Bloom Lab.
Dr. Russell and Dr. Jesse Bloom, in collaboration with Dr. Cole Trapnell in the UW Department of Genome Sciences, performed single-cell mRNA sequencing to measure the abundance of all cellular and viral transcripts within host cells exposed to a low infectious dose of influenza virus. Their results, recently published in eLife, reveal a startling level of heterogeneity in the amount of viral RNA present in each infected cell.
To begin, Dr. Russell exposed a population of human cells to influenza virus for different amounts of time and then used an oil-water emulsion-based system to physically isolate single cells inside stable droplets. By introducing barcoded primers into each droplet, all transcripts originating from the same cell were tagged with a unique sequence during reverse transcription of RNA into cDNA. In addition, each primer molecule contained a unique molecular identifier (UMI) to allow for elimination of PCR amplification bias, a common problem in high-throughput sequencing. Once all transcripts were tagged with cell-specific barcodes and UMIs, they were pooled and sequenced at the 3’ end to determine transcript identity.
The researchers obtained sequencing data for 3,000-4,000 cells per sample. After applying rigorous selection criteria to differentiate truly infected cells from uninfected cells that obtained viral mRNA via leakage from lysed cells, the authors determined that there were 50-150 infected cells per sample, ~10% of which were originally co-infected by more than one viral particle. Among infected cells, the fraction of mRNA derived from influenza was highly variable (from <1% to >50%) and followed an exponential distribution, with most cells containing low amounts of viral RNA and a few containing very high levels. “This result indicates that the successful production of viral material in cells is very unevenly distributed, even more so than income inequality in the United States,” says Dr. Bloom.
Next, the researchers sought to identify the causes of this variation. Viral mRNA levels did not track with how long cells had been infected or whether they were co-infected, suggesting that time since initiation of infection and infectious dose are not major contributors.
Because the influenza genome is encoded on eight separate DNA segments and viral packaging is known to be error-prone, some virions were likely to be missing at least one genome segment. Indeed, the authors observed that one or more viral genes failed to be expressed in about half of infected cells. Infected cells lacking any one of the four segments required for producing the viral ribonucleoprotein (RNP) complex never had high viral RNA levels, consistent with the known dependence of viral transcription on RNP. However, the amount of viral RNA still varied widely among cells expressing all viral genes, indicating that there must be additional sources of variability.
The researchers were surprised to observe that, despite the wide variation in the amount of viral RNA per cell, the relative amounts of each viral transcript within a given cell were remarkably similar. This result suggests that influenza has mechanism(s) for maintaining relative expression of its genome segments. In addition, the data revealed that a co-infecting virus can complement another virion’s missing genome segment, thus constituting the first direct observation of this phenomenon in single cells.
Since the sequencing results included both viral and cellular RNAs, the authors could also analyze host transcription during infection. Prior studies had observed induction of interferon, a component of the innate immune system, in response to large infectious doses. However, only one cell in Dr. Russell’s samples expressed detectable interferon mRNAs, likely because the low rate of infection was insufficient to induce immune signaling between neighboring cells and because the virus population used in the infections contained fewer defective particles compared to previous studies. Interestingly, the virus infecting the interferon-expressing cell lacked NS1, a viral protein known to antagonize the interferon response, although other viruses lacking NS1 did not induce interferon.
Finally, the authors sought to identify correlations between host transcription and viral RNA levels. They noticed that genes involved in the response to reactive oxygen species or hypoxia were significantly higher in cells with more viral mRNA. Oxidative stress is known to occur during influenza infection and is thought to be beneficial to the virus—and possibly even actively induced by it. In accordance with this notion, follow-up experiments performed by Dr. Russell revealed that cells pre-treated with an antioxidant were less likely to express detectable levels of the viral surface protein HA.
Overall, this study reveals a striking heterogeneity in how well individual viral particles can convert host cells into virus-making factories, an insight that could not have been made by making bulk measurements. "This work shows how remarkably different viral infection looks if you examine individual cells rather than large populations,” says Dr. Bloom. As stated by Dr. Russell, “we hope that our findings and methodology will serve as a jumping-off point for further studies and for a shift in how viral disease is studied.”
Russell AB, Trapnell C and Bloom JD. 2018. Extreme heterogeneity of influenza virus infection in single cells. eLife.
This research was supported by the Damon Runyon Cancer Research Foundation, the Burroughs Wellcome Fund, the Simons Foundation and the Howard Hughes Medical Institute.