Innate Immune Reconstitution in the Humanized Bone Marrow-Liver-Thymus (HuBLT) Mouse Model Is Essential for Adaptive Immune Responses to HIV-1 Infection and Can Be Enhanced via AAV-Mediated Human Cytokine Delivery

BACKGROUND: Mice harboring a human immune system (a.k.a. “humanized” mice) are a revolutionary small-animal model consisting of a human-to-mouse hematopoietic xenograft that allows for the scientific and clinical study of human immune development and function, therapeutic agents, vaccines, and pathogens. In particular, it has served as an exceptional model of infection with human immunodeficiency virus type 1 (HIV-1), a human-restricted pathogen whose closest animal model is non-human primates infected with simian immunodeficiency virus (SIV). Of the existing humanized mouse models, humanized bone marrow-liver-thymus (HuBLT) mice are considered one of the most advanced, as they uniquely harbor development of human T cells in vivo in an autologous human thymic graft. Several groups have shown that HuBLT mice recapitulate many aspects of acute and chronic HIV-1 infection, and have greatly aided in advancing research focused on areas such as antiretroviral therapy, broadly neutralizing antibodies, viral evolution, and vaccine development. However, HuBLT mice challenged with HIV-1 exhibit variability in human immune responses across and within groups of engrafted mice, hindering our ability to confidently detect HIV-1–specific responses or vaccine effects in small cohorts of mice. To understand the phenomena underlying this variability, we comprehensively analyzed T-cell development, diversity, and priming in HuBLT mice to identify any model-intrinsic defects that could be corrected. METHODS AND RESULTS: Through the use of TCR sequencing, flow cytometric analyses, and cellular immunological assays, we found that while T-cell diversity generation, thymic development, and subset frequencies were grossly intact, there was a major defect in T-cell priming and function. This defect correlated with poor innate immune reconstitution in HuBLT mice. We found that while almost all HuBLT mice reconstituted well with CD4+ and CD8+ T cells, only the few mice that reconstituted a substantial amount of monocytes (≥1% CD14+ cells of human CD45+ cells) had robust CD4+ and CD8+ T-cell responses and responded appropriately during acute HIV-1 infection, as determined by CD4+ T-cell decline and CD8+ T-cell activation, expansion, and differentiation. This suggested that T-cell priming in the HuBLT mice was exquisitely sensitive to the presence and frequency of innate immune cells (e.g. monocytes) that serve as antigen-presenting cells. Thus, deficient innate immune reconstitution was found to be a key contributor to the variability of anti-HIV-1 immune responses seen in HuBLT mice studies. Given that sub-optimal innate immune reconstitution was the major barrier to proper immune response priming in HuBLT mice, we sought to correct this. Several studies have postulated that lack of cross-reactivity between human and mouse cytokines and growth factors involved in hematopoiesis hinder myelopoiesis more than lymphopoiesis in humanized mouse models. Investigators have made efforts to overcome this by exogenous administration, hydrodynamic transfection, or genetic engineering of mouse strains to supply these human factors. As these cytokine-enhancement modalities are expensive, cumbersome, and/or time-intensive, we developed an in-vivo transduction strategy using adeno-associated virus (AAV) vectors. We generated a library of AAVs encoding a multitude of different human cytokines (AAV-hCYTs) that can be delivered in a single administration to humanized mice (singly or in cocktails) to provide tunable, long-lasting expression of any protein(s) of interest, which we refer to as ‘AAV-mediated cytokine enhancement’ (ACE). This allowed us to test different human cytokines at specific doses and assess their effect on the engraftment and functionality of different human immune cell subsets, which we characterized by flow cytometric analysis of peripheral blood and tissues (i.e. spleen) as well as multiplexed ELISAs (i.e. Luminex) of human cytokines in mouse plasma. Single AAV-hCYTs in HuBLT mice showed that permanent perturbations of the human immune system were possible. For example, IL-15 expanded NK cells; IL-2 increased Tregs (CD4+CD25+CD127– T cells) and NK cells; and GM-CSF increased the frequency of classical (CD14+CD16–) monocytes, Tregs, and memory T-cell subsets, which were paralleled by increases in corresponding plasma cytokines (e.g. CCL2, IL-10, and CCL4). Given these results, we generated HuBLT mice expressing cocktails of AAV-hCYTs. While some cocktails were resulted in morbidity due to immune over-activation and cytokine storm, others showed remarkable improvements in innate immune reconstitution and adaptive immune function. For example, mice expressing AAV-delivered SCF, GM-CSF, and IL-3 had increased frequencies of myeloid-origin innate immune cells (CD11c+ cells) in peripheral blood, which was associated with increased frequencies of memory T-cell subsets and Tregs. These mice also exhibited increased plasma levels of IgG1 and IgG3 (~50-fold and ~2.8-fold more, respectively), indicating a positive effect on B-cell function. In addition, HIV-1 infection of these mice showed more consistent and uniform viremia kinetics. These data thus show ACE can overcome model-intrinsic limitations and improve human engraftment and immune responses in HuBLT mice. CONCLUSION: While further efforts to generate personalized AAV cocktails for manipulation of human immune reconstitution and function in HuBLT mice are ongoing, it is clear that AAV-mediated delivery of human cytokines are an efficient way to generate a more optimal humanized mouse model. Given its flexibility, tunability, and portability, we believe the ACE platform can aid in exploring different facets of in-vivo human immunology and anti-viral immune responses in ways that were not previously possible