Quantitative imaging and dynamics of tumor therapy with viruses.

Cancer therapy remains challenging due to the myriad presentations of the disease and the vast genetic diversity of tumors that continuously evolve and often become resistant to therapy. Viruses can be engineered to specifically infect, replicate, and kill tumor cells (tumor virotherapy). Moreover, the viruses can be "armed" with therapeutic genes to enhance their oncolytic effect. Using viruses to treat cancer is exciting and novel and in principle can be used for a broad variety of tumors. However, the approach is distinctly different from other cancer therapies since success depends on establishment of an infection within the tumor and ongoing propagation of the oncolytic virus within the tumor itself. Therefore, the target itself amplifies the therapy. This introduces complex dynamics especially when the immune system is taken into consideration as well as the physical and other biological barriers to virus growth. Understanding these dynamics not only requires mathematical and computational models but also approaches for the noninvasive monitoring of the virus and tumor populations. In this perspective, we discuss strategies and current results to achieve this important goal of understanding these dynamics in pursuit of optimization of oncolytic virotherapy.

Preclinical development of CD126 CAR-T cells with broad antitumor activity.

Chimeric antigen receptor T (CAR-T) cell therapy is a transformative approach to cancer eradication. CAR-T is expensive partly due to the restricted use of each CAR construct for specific tumors. Thus, a CAR construct with broad antitumor activity can be advantageous. We identified that CD126 is expressed by many hematologic and solid tumors, including multiple myeloma, lymphoma, acute myeloid leukemia, pancreatic and prostate adenocarcinoma, non-small cell lung cancer, and malignant melanoma among others. CAR-T cells targeting CD126 were generated and shown to kill many tumor cells in an antigen-specific manner and with efficiency directly proportional to CD126 expression. Soluble CD126 did not interfere with CAR-T cell killing. The CAR-T constructs bind murine CD126 but caused no weight loss or hepatotoxicity in mice. In multiple myeloma and prostate adenocarcinoma xenograft models, intravenously injected CD126 CAR-T cells infiltrated within, expanded, and killed tumor cells without toxicity. Binding of soluble interleukin-6 receptor (sIL-6R) by CAR-T cells could mitigate cytokine release syndrome. Murine SAA-3 levels were lower in mice injected with CD126 CAR-T compared to controls, suggesting that binding of sIL-6R by CAR-T cells could mitigate cytokine release syndrome. CD126 provides a novel therapeutic target for CAR-T cells for many tumors with a low risk of toxicity.

In vitro and in silico multidimensional modeling of oncolytic tumor virotherapy dynamics.

Tumor therapy with replication competent viruses is an exciting approach to cancer eradication where viruses are engineered to specifically infect, replicate, spread and kill tumor cells. The outcome of tumor virotherapy is complex due to the variable interactions between the cancer cell and virus populations as well as the immune response. Oncolytic viruses are highly efficient in killing tumor cells in vitro, especially in a 2D monolayer of tumor cells, their efficiency is significantly lower in a 3D environment, both in vitro and in vivo. This indicates that the spatial dimension may have a major influence on the dynamics of virus spread. We study the dynamic behavior of a spatially explicit computational model of tumor and virus interactions using a combination of in vitro 2D and 3D experimental studies to inform the models. We determine the number of nearest neighbor tumor cells in 2D (median = 6) and 3D tumor spheroids (median = 16) and how this influences virus spread and the outcome of therapy. The parameter range leading to tumor eradication is small and even harder to achieve in 3D. The lower efficiency in 3D exists despite the presence of many more adjacent cells in the 3D environment that results in a shorter time to reach equilibrium. The mean field mathematical models generally used to describe tumor virotherapy appear to provide an overoptimistic view of the outcomes of therapy. Three dimensional space provides a significant barrier to efficient and complete virus spread within tumors and needs to be explicitly taken into account for virus optimization to achieve the desired outcome of therapy.

In Vivo Imaging of Oncolytic Measles Virus Propagation with Single-Cell Resolution.

Recombinant measles viruses (MVs) have oncolytic activity against a variety of human cancers. However, their kinetics of spread within tumors has been unexplored. We established an intravital imaging system using the dorsal skin fold chamber, which allows for serial, non-invasive imaging of tumor cells and replication of a fusogenic and a hypofusogenic MV. Hypofusogenic virus-infected cells were detected at the earliest 3 days post-infection (dpi), with peak infection around 6 dpi. In contrast, the fusogenic virus replicated faster: infected cells were detectable 1 dpi and cells were killed quickly. Infection foci were significantly larger with the fusogenic virus. Both viruses formed syncytia. The spatial relationships between cells have a major influence on the outcome of therapy with oncolytic viruses.

In Vivo Estimation of Oncolytic Virus Populations within Tumors.

The use of replication-competent viruses as oncolytic agents is rapidly expanding, with several oncolytic viruses approved for cancer therapy. As responses to therapy are highly variable, understanding the dynamics of therapy is critical for optimal application of virotherapy in practice. Although mathematical models have been developed to understand the dynamics of tumor virotherapy, a scarcity of in vivo data has made difficult parametrization of these models. To tackle this problem, we studied the in vitro and in vivo spread of two oncolytic measles viruses that induce expression of the sodium iodide symporter (NIS) in cells. NIS expression enabled infected cells to concentrate radioactive isotopes that could be reproducibly and quantitatively imaged using SPECT/CT. We observed a strong linear relationship in vitro between infectious virus particles, viral N and NIS gene expression, and radioactive isotope uptake. In vivo radioisotope uptake was highly correlated with viral N and NIS gene expression. Similar expression patterns between viral N and NIS gene expression in vitro and in vivo implied that the oncolytic virus behaved similarly in both scenarios. Significant titers of viable virus were consistently isolated from tumors explanted from mice that had been injected with oncolytic measle viruses. We observed a weaker but positive in vivo relationship between radioisotope uptake and the viable virus titer recovered from tumors; this was likely due to anisotropies in the viral distribution in vivo These data suggest that methods that enable quantitation of in vivo anisotropies are required for continuing development of oncolytic virotherapy.Significance: These findings address a fundamental gap in our knowledge of oncolytic virotherapy by presenting technology that gives insight into the behavior of oncolytic viruses in vivo Cancer Res; 78(20); 5992-6000. ©2018 AACR.

Human CRM1 augments production of infectious human and feline immunodeficiency viruses from murine cells.

Productive replication of human immunodeficiency virus type 1 (HIV-1) occurs efficiently only in humans. The posttranscriptional stages of the HIV-1 life cycle proceed poorly in mouse cells, with a resulting defect in viral assembly and release. Previous work has shown that the presence of human chromosome 2 increases HIV-1 production in mouse cells. Recent studies have shown that human chromosome region maintenance 1 (hCRM1) stimulates Gag release from rodent cells. Here we report that expressions of hCRM1 in murine cells resulted in marked increases in the production of infectious HIV-1 and feline immunodeficiency virus (FIV). HIV-1 production was also increased by hSRp40, and a combination of hCRM1 and hSRp40 resulted in a more-than-additive effect on HIV-1 release. In contrast, the overexpression of mouse CRM1 (mCRM1) minimally affected HIV-1 and FIV production and did not antagonize hCRM1. In the presence of hCRM1 there were large increases in the amounts of released capsid, which paralleled the increases in the infectious titers. Consistent with this finding, the ratios of unspliced to spliced HIV-1 mRNAs in mouse cells expressing hCRM1 and SRp40 became similar to those of human cells. Furthermore, imaging of intron-containing FIV RNA showed that hCRM1 increased RNA export to the cytoplasm.By testing chimeras between mCRM1 and hCRM1 and comparing those sequences to feline CRM1, we mapped the functional domain to HEAT (Huntingtin, elongation factor 3, protein phosphatase 2A, and the yeast kinase TOR1) repeats 4A to 9A and a triple point mutant in repeat 9A, which showed a loss of function. Structural analysis suggested that this region of hCRM1 may serve as a binding site for viral or cellular factors to facilitate lentiviral RNA nuclear export.

Feline immunodeficiency virus Gag is a nuclear shuttling protein.

Lentiviral genomic RNAs are encapsidated by the viral Gag protein during virion assembly. The intracellular location of the initial Gag-RNA interaction is unknown. We previously observed feline immunodeficiency virus (FIV) Gag accumulating at the nuclear envelope during live-cell imaging, which suggested that trafficking of human immunodeficiency virus type 1 (HIV-1) and FIV Gag may differ. Here we analyzed the nucleocytoplasmic transport properties of both Gag proteins. We discovered that inhibition of the CRM1 nuclear export pathway with leptomycin B causes FIV Gag but not HIV-1 Gag to accumulate in the nucleus. Virtually all FIV Gag rapidly became intranuclear when the CRM1 export pathway was blocked, implying that most if not all FIV Gag normally undergoes nuclear cycling. In FIV-infected feline cells, some intranuclear Gag was detected in the steady state without leptomycin B treatment. When expressed individually, the FIV matrix (MA), capsid (CA), and nucleocapsid-p2 (NC-p2) domains were not capable of mediating leptomycin B-sensitive nuclear export of a fluorescent protein. In contrast, CA-NC-p2 did mediate nuclear export, with MA being dispensable. We conclude that HIV-1 and FIV Gag differ strikingly in a key intracellular trafficking property. FIV Gag is a nuclear shuttling protein that utilizes the CRM1 nuclear export pathway, while HIV-1 Gag is excluded from the nucleus. These findings expand the spectrum of lentiviral Gag behaviors and raise the possibility that FIV genome encapsidation may initiate in the nucleus.

Live-cell coimaging of the genomic RNAs and Gag proteins of two lentiviruses.

Human immunodeficiency virus type 1 (HIV-1) Gag and genomic RNA determinants required for encapsidation are well established, but where and when encapsidation occurs in the cell is unknown. We constructed MS2 phage coat protein labeling systems to track spatial dynamics of primate and nonprimate lentiviral genomic RNAs (HIV-1 and feline immunodeficiency virus [FIV]) vis-à-vis their Gag proteins in live cells. Genomic RNAs of both lentiviral genera were observed to traffic into the cytoplasm, and this was Rev dependent. In transit, FIV Gag and genomic RNA accumulated independently of each other at the nuclear envelope, and focal colocalizations of genomic RNA with an intact packaging signal (psi) and Gag were observed to extend outward from the cytoplasmic face. In contrast, although HIV-1 genomic RNA was detected at the nuclear envelope, HIV-1 Gag was not. For both lentiviruses, genomic RNAs were seen at the plasma membrane if and only if Gag was present and psi was intact. In addition, HIV-1 and FIV genomes accumulated with Gag in late endosomal foci, again, only psi dependently. Thus, lentiviral genomic RNAs require specific Gag binding to accumulate at the plasma membrane, packaged genomes cointernalize with Gag into the endosomal pathway, and plasma membrane RNA incorporation by Gag does not trigger committed lentiviral particle egress from the cell. Based on the FIV results, we hypothesize that the Gag-genome association may initiate at the nuclear envelope.

The critical role of proximal gag sequences in feline immunodeficiency virus genome encapsidation.

Retroviral RNA encapsidation is mediated by specific interactions between viral Gag proteins and cis-acting packaging sequences in genomic RNA. Feline immunodeficiency virus (FIV) RNA encapsidation determinants have been shown to be discrete and noncontinuous, comprising one region at the 5' end of the genomic mRNA (R-U5) and another region that mapped within the proximal 311 nt of gag. To aid comparative understanding of lentiviral encapsidation and refinement of FIV vector systems, we used RNase protection assays (RPAs) of cellular and virion RNAs to investigate in detail the gag element. mRNAs of subgenomic vectors as well as of full-length molecular clones were optimally packaged into viral particles and resulted in high-titer FIV vectors when they contained only the proximal 230 nucleotides (nt) of gag. Further 3' truncations of gag sequences progressively diminished encapsidation and transduction. Deletion of the initial ninety 5' nt of the gag gene abolished mRNA packaging, demonstrating that this segment is indispensable for encapsidation. Focusing further on this proximal sequence, we found that a deletion of only 13 nt at the 5' end of gag impaired encapsidation of subgenomic vector and proviral RNAs.

Mapping the encapsidation determinants of feline immunodeficiency virus.

Encapsidation of retroviral RNA involves specific interactions between viral proteins and cis-acting genomic RNA sequences. Human immunodeficiency virus type 1 (HIV-1) RNA encapsidation determinants appear to be more complex and dispersed than those of murine retroviruses. Feline lentiviral (feline immunodeficiency virus [FIV]) encapsidation has not been studied. To gain comparative insight into lentiviral encapsidation and to optimize FIV-based vectors, we used RNase protection assays of cellular and virion RNAs to determine packaging efficiencies of FIV deletion mutants, and we studied replicative phenotypes of mutant viruses. Unlike the case for other mammalian retroviruses, the sequences between the major splice donor (MSD) and the start codon of gag contribute negligibly to FIV encapsidation. Moreover, molecular clones having deletions in this region were replication competent. In contrast, sequences upstream of the MSD were important for encapsidation, and deletion of the U5 element markedly reduced genomic RNA packaging. The contribution of gag sequences to packaging was systematically investigated with subgenomic FIV vectors containing variable portions of the gag open reading frame, with all virion proteins supplied in trans. When no gag sequence was present, packaging was abolished and marker gene transduction was absent. Inclusion of the first 144 nucleotides (nt) of gag increased vector encapsidation to detectable levels, while inclusion of the first 311 nt increased it to nearly wild-type levels and resulted in high-titer FIV vectors. However, the identified proximal gag sequence is necessary but not sufficient, since viral mRNAs that contain all coding regions, with or without as much as 119 nt of adjacent upstream 5' leader, were excluded from encapsidation. The results identify a mechanism whereby FIV can encapsidate its genomic mRNA in preference to subgenomic mRNAs.