Silence
is a Behaviour
Quantitative Virology Research Group
One of the main objectives of our laboratory is to study the establishment of HIV latency coupled with stochastic HIV transcription and the host functional genome.
Using quantitative gnomic and machine-learning-based approaches tailored from the laboratory, we aim to deepen our grasp on how variations, with resolution of single viruses, in HIV genome integrity and transcription influence the configuration of HIV reservoirs. Two research lines aligning with this objective are conducted.
1. Role of HIV antisense transcripts (AST) in the establishment of latency
We propose that fluctuations in HIV transcription can appear at, at least two levels: (1) the chromosomal landscape and (2) the in situ HIV integration site (Figure 1). In the former case, proviruses integrating in different genomic locations demonstrate a variety of transcriptional bursting and can be classified in noise space constructed by parameters associated with coefficient of variation or using a mathematical model fitting a curve of exponential decay. In the latter case, stochastic HIV gene expression tends to be a pure epigenetic phenomenon: the identical provirus demonstrates the turnover of HIV infection with elevated frequency. We observed inheritance of such stochastic fluctuations and identified the provirus integration site, which is in the vicinity of the genomic repetitive regions associated with intense signals of H3K27ac and H3K4me3. In both cases, we observed similar consecutive expression patterns and a positive correlation between sense and antisense RNA transcripts over time, suggesting that the independent yields of both transcripts determine the stochastic phenotype of provirus transcription, rather than their ratios. Notably, both HIV LTRs are responsive to drug stimulation and could differently react to drugs as transcriptional bursting is epigenetics-driven. Overall, our data suggest that antisense transcripts are most likely involved in the stochastic nature of HIV transcription.
2. Distinguishable topology of the task-evoked functional genome networks in HIV reservoirs
Our len to view HIV reservoir is that it can be represented as the topological (task-evoked) property of a network consisting of different communities of the genes targeted by HIV. Such communities are so-called immunologic signatures. HIV integration frequency within a network might be used as a proxy to define specific immune cell types and proinflammatory soluble factors, facilitating fine-tuning of the microenvironment of reservoirs. To deepen our understanding of such heterogeneous HIV reservoirs and their functional implications, we pioneer the integration of a convergence approach to characterize the composition of HIV reservoirs.
Based on graph-theoretical tools, we observe noticeable topological properties in networks, featuring immunologic signatures enriched by genes harboring intact and defective proviruses, when comparing antiretroviral therapy (ART)-treated HIV-1-infected individuals and elite controllers. The key variable, the rich factor, plays a pivotal role in classifying distinct topological properties in networks. The host gene expression strengthens the accuracy of classification between elite controllers and ART-treated patients. Markov chain Monte Carlo modeling for the simulation of different graph networks demonstrated the presence of an intrinsic barrier between elite controllers and non-elite controllers. Notably, our work provides a prime example of leveraging genomic approaches alongside mathematical tools to unravel the complexities of HIV reservoirs.

Figure 2. The network topology of HIV reservoirs in ART-treated patients versus elite controllers. In comparison to elite controllers, the network architecture in ART-treated patients exhibits three key characteristics (1) a less intense magnitude of signature enrichment, (2) a high degree of assortativity, and (3) elevated connectedness between two adjacent vertices. These findings suggest that the network architecture is more connective and structural in ART-treated patients.
Publications related to this research line:
1. Chen, H.-C. 2023. Vaccines DOI:10.3390/vaccines11020402
2. Więcek, K., and Chen, H.-C. 2023. iScience DOI:10.1016/j.isci.2023.108342
3. Wiśniewski, et al. 2024. iScience DOI:10.1016/j.isci.2024.11122

Figure 1. The involvement of HIV asRNAs in stochastic fluctuations in HIV gene expression. We propose that fluctuations in HIV transcription can appear at, at least two levels: (1) the chromosomal landscape and (2) the HIV integration site (A, B). We observed a minimum level of inheritance of such stochastic fluctuations and identified the provirus integration site, which is in the vicinity of the genomic repetitive regions associated with intense signals of H3K27ac and H3K4me3. In both cases, similar expression patterns between sense and antisense RNA transcripts in a time series were observed (C), suggesting that the independent yields of both transcripts determine the stochastic phenotype of provirus transcription, rather than their ratios.
Publications related to this research line:
Więcek, K. et al. 2025
Another scientific objective in the laboratory is to study the linkage between the intrinsic tropisms of coronavirus variants and host domestication.
3. Identification of potential SARS-CoV-2 genetic markers resulting from host domestication
We develop a k-mer-based pipeline, namely the Pathogen Origin Recognition Tool using Enriched K-mers (PORT-EK) to identify genomic regions enriched in the respective hosts after the comparison of metagenomes of isolates between two host species. Using it we successfully identify thousands of k-mers enriched in US white-tailed deer and betacoronaviruses while comparing them with human isolates. In addition, we demonstrate different coverage landscapes of k-mers enriched in deer and bats and unravel 144 mutations in enriched k-mers yielded from the comparison of viral metagenomes between bat and human isolates. Additionally, we observe that the third position within a genetic codon is prone to mutations, resulting in a high frequency of synonymous mutations of amino acids harboring the same physicochemical properties as unaltered amino acids. Importantly, we are able to classify and predict the likelihood of host species based on the enriched k-mer counts.

Figure 3. Rational design of PORT-EK and determination of the enriched k-mers. (A) The analytical pipeline of PORT-EK. PORT-EK consists of four steps including (1) k-mers matrix preparation, (2) k-mers filtering and selection, (3) the identification of host-specific mutations, and (4) the classification of hosts. (B) Funnel plot representing the filtering strategies for the selection of enriched k-mers. Four layers of filtering are applied in the PORT-EK pipeline.
Publications related to this research line:
1. Wiśniewski and Chen. 2024. BioRxiv DOI:10.1101/2024.07.27.605454