Machine learning prediction and phyloanatomic modeling of viral neuroadaptive signatures in the macaque model of HIV-mediated neuropathology
In human immunodeficiency virus (HIV) infection, virus replication in the central nervous system (CNS) can result in HIV-associated neurocognitive deficits in approximately 25% of patients with unsuppressed viremia and is thought to be characterized by evolutionary adaptation to this unique microenvironment. While no single mutation can be agreed upon as distinguishing the neuroadapted population from virus in patients without neuropathology, earlier studies have demonstrated that a machine learning (ML) approach could be applied to identify a collection of mutational signatures within the envelope glycoprotein (Env Gp120) predictive of disease. The S[imian] IV-infected macaque is a widely used animal model of HIV neuropathology, allowing in-depth tissue sampling infeasible for human patients. Yet, translational impact of the ML approach within the context of the macaque model has not been tested, much less the capacity for early prediction in other, non-invasive tissues. We applied the previously described ML approach to prediction of SIV-mediated encephalitis (SIVE) using gp120 sequences obtained from the CNS of animals with and without SIVE with 73% accuracy. The presence of SIVE signatures at earlier time points of infection in non-CNS tissues in both SIVE and SIVnoE animals indicated these signatures cannot be used in a clinical setting. However, combined with protein structural mapping and statistical phylogenetic inference, results revealed common denominators associated with these signatures, including 2-acetamido-2-deoxy-beta-D-glucopyranose structural interactions and the infection of alveolar macrophages. Alveolar macrophages were demonstrated to harbor a relatively large proportion (35 - 100%) of SIVE-classified sequences and to be the phyloanatomic source of cranial virus in SIVE, but not SIVnoE animals. While this combined approach cannot distinguish the role of this cell population as an indicator of cellular tropism from a source of neuroadapted virus, it provides a key to understanding the function and evolution of the signatures identified as predictive of both HIV and SIV neuropathology. Author summaryHIV-associated neurocognitive disorders remain prevalent among HIV-infected individuals, even in the era of potent antiretroviral therapy, and our understanding of the mechanisms involved in disease pathogenesis, such as virus evolution and adaptation, remains elusive. In this study, we expand on a machine learning method previously used to predict neurocognitive impairment in HIV-infected individuals to the macaque model of AIDS-related neuropathology in order to characterize its translatability and predictive capacity in other sampling tissues and time points. We identified four amino acid and/or biochemical signatures associated with disease that, similar to HIV, demonstrated a proclivity for proximity to aminoglycans in the protein structure. These signatures were not, however, isolated to specific points in time or even to the central nervous system, as they could be observed at low levels during initial infection and from various tissues, most prominently in the lungs. The spatiotemporal patterns observed limit the use of these signatures as an accurate prediction for neuropathogenesis prior to the onset of symptoms, though results from this study warrant further investigation into the role of these signatures, as well as lung tissue, in viral entry to and replication in the brain.