Vesicular stomatitis virus causes lethal neurotoxicity in mice.
Vesicular stomatitis virus is capable of causing lethal encephalitis in rodents via almost any administration route. Direct intracranial injection of the virus, however, results in development of neuropathogenesis faster and at a lower dose than infections initiated at other sites. In order to provide the most stringent evaluation of viral restriction in the brain, we first analyzed the propensity of an engineered VSV containing a firefly luciferase reporter (VSV-Luc) to cause lethal encephalitis when the virus was administered intracerebrally. We first investigated the time course of gene expression in immunocompromised mice inoculated with 1 × 104 VSV-Luc particles and examined the luciferase expression kinetics post-viral administration. Not surprisingly, luciferase expression (a marker for viral gene expression) was visible in the brain at 24 h postadministration, and expression continued to increase until animals were euthanized or succumbed to neurotoxicity. Often, the onset of paralysis occurred at the time that luciferase was visualized in the spinal cords of infected mice (Fig. ). Mice inoculated with VSV-Luc intracranially exhibited classic signs of neurotoxicity, including listing, seizure, weight loss, and paralysis, and died or were euthanized prior to 15 days postinoculation. Upon histological examination of the brain, control Opti-MEM-treated mice had no histopathology outside the needle tract after administration of control carrier (Fig. ), while mice treated with VSV-Luc had numerous inflammatory lesions and moderate to severe meningoencephalitis, as visualized by hematoxylin and eosin (H&E) staining (Fig. ).
FIG. 1. Vesicular stomatitis virus causes lethal neurotoxicity. (A) VSV-Luc localizes to and replicates in the brains of mice. Control Opti-MEM carrier or 1 × 104 VSV-Luc particles were injected intracranially into SCID mice and imaged for luminescence (more ...)
In order to determine the 50% lethal dose (LD50) of the parental VSV-Luc that would then be modified by miRNA target (miRT) insertion, we did a classic viral dose escalation study. Virus doses ranged from 1 × 102 to 1 × 107 in 10 μl of carrier. Not surprisingly, at the top doses of 1 × 104 to 1 × 107, all animals developed neurotoxicity and died or were euthanized prior to 15 days postinoculation (Fig. ). Animals that were administered less than 1,000 infectious units, however, were spared from lethal encephalitis, and the LD50 was determined to be 5,000 TCID50.
Neuronal microRNAs downregulate gene expression in lentiviral vectors.
Over 700 miRNAs have been identified in humans (35
), many of which are known to have tissue-specific expression profiles (31
). As we were interested in ameliorating viral encephalitis, we investigated those miRNAs that are known to be abundant in the CNS and which would be absent in tumors. Numerous brain-specific miRNAs have been identified (44
) and are known to be expressed in different abundances, locations, and cell lineages. To identify the target elements that would most efficiently shut down gene expression in the brain, we first employed lentiviral vectors containing different brain-specific miRTs.
To select the optimal sequences for brain-specific shutdown, we utilized our previous reporter system whereby four tandem miRTs are inserted downstream of a spleen focus-forming virus (SFFV)-driven reporter gene in a lentiviral vector (Fig. ). We chose four of the best-characterized brain-specific target elements, as well as a hematopoiesis-specific miRT control, for further investigation. The most abundant miRNA in the brain is miR-124, a neuron-specific miRNA that has a role in neural specification in the developing brain. Vectors containing target elements corresponding to miR-124 were constructed, as well as vectors expressing miR-125T, which is expressed ubiquitously in all cell lineages of the brain; miR-128T, which has previously been shown to silence gene expression in neurons (48
); and the recently identified miR-134T, which regulates spine development in addition to being heavily expressed in the hippocampus (Fig. ).
Vectors expressing control and brain-specific miRTs were transduced on a panel of cell lines thought to be indicative of different cell lineages that would have different miRNA expression profiles (Fig. ). Since neuronal cells express all four miRNAs (miR-124, miR-125, miR-128, and miR-134), we first transduced cells of different neuronal origins with our miRNA sensor vectors: Cath.A and NB41A3 are neuroblastomas, while 293T cells (once thought to be epithelial, but now reclassified) are transformed neuronal cells (45
). While vectors containing target elements for all neuronal targets assayed did show significant decreases in gene expression in cells of neuronal lineage compared to HeLa and Mel-624 control cells (P
< 0.05), the decrease in most cases was no greater than 30% (Fig. ). Even when miRT sensors were transduced into cells that contained fully complementary miRNA mimics in trans
, a decrease in luciferase expression was not seen with all targets. Only when a vector containing miR-125T, miR-134T, or a combination of these targets was present was gene expression reduced by over 90% (Fig. ). While we (26
) and others (8
) have shown previously that combining more than one miRT can have additive affects in decreasing gene expression, we did not see this with neuronal targets (Fig. and data not shown).
The factors affecting the efficiency with which miRNAs are able to silence sequence-complementary mRNAs are very poorly understood. While it has been shown in certain cases that more abundant miRNAs functionally silence miRTs better than do low-abundance miRNAs (8
), it is clear that expression level is not the only factor that affects the regulatory capacity of miRNAs (6
VSV accommodates miRT insertion without slowing of growth kinetics.
Based upon our data obtained from lentiviral vectors as well as its expression profile in the brain, we determined that miR-125 provided the best candidate for attenuating viral gene expression in the brain. Vectors containing miR-125T showed a >95% decrease in luciferase expression in the presence of fully complementary miRNA mimics in vitro
(Fig. ). In addition, all cell lineages in the brain (neurons, astrocytes, and glia cells) are thought to express miR-125 (though at different abundances). We next constructed fully replication-competent rVSVs that expressed target elements complementary to miR-125. As a control, we employed targets corresponding to miR-206, which we have shown previously to silence gene expression in muscle but to provide no protection from neurovirulence (26
Endogenous miRTs are most often located in the 3′UTRs of cellular mRNAs. While some targets are found in coding regions of mRNAs as well as the 5′UTR, it is thought that there is a particular bias for efficacy of miRT localization in the 3′UTR (19
). VSV, unlike picornaviruses, expresses each viral protein from a distinct mRNA. Unlike the design of miRNA-responsive picornaviruses, VSV provided a much greater burden for miRNA regulation, since each miRT would target only that gene in which the target was placed.
VSV contains five different viral genes, with transcriptional stop sites after each gene. Intergenic gaps function as a signal for polyadenylation and mRNA termination, and then conserved transcriptional start sites start transcription of the next gene (28
). Due to transcriptase drop-off at each site, VSV genes are expressed in a gradient, with the proximal viral N gene having the most transcripts and the distal L gene having the least. This transcriptional gradient is important, as a high expression level of L is very cytotoxic and N must be expressed at sufficient levels to wrap the viral genome into a ribonucleoprotein (RNP) complex and also to provide sufficient protein to encapsidate the virus. The expression levels of M, P, and G appear to be less important, as rearrangement of these viral genes has been shown to be of little consequence to viral replication (3
Bearing in mind that the viral L protein is expressed at the lowest levels of all viral genes, we hypothesized that it might be particularly vulnerable to miRNA-mediated inhibition. Also, while the expression level of the viral M protein is not as important in vitro, the M gene is responsible for the inhibition of host mRNA export that disables the immune response against the virus. Therefore, because the viral matrix protein inhibits an antiviral cellular signal cascade, we thought it could also be a target of potential importance. In addition to the M and L genes, we included target elements in the luciferase gene of the virus as a control. Viruses containing miRNA-targeted luciferase while potentially having less reporter should not have affected viral life cycles. Targets were therefore inserted into the 3′UTRs of the viral M, Luc, and L genes (Fig. ). In an effort to target the entire antigenome, we also inserted miRT sites within the viral 3′UTR, but these insertions prohibited viral rescue.
FIG. 3. Construction and characterization of recombinant VSVs. (A) Schematic diagram of microRNA-targeted VSV. A novel NotI restriction site was created in the 3′UTR of the M, Luc, or L gene, and miRTs were cloned into this site. (B) One-step growth curves (more ...)
Since VSV is a minus-strand rhabdovirus, the VSV genome and mRNA are expressed in different orientations. We have shown previously that direct targeting of the viral genomes of picornaviruses provides the most efficient miRNA-mediated inhibition (Kelly et al., unpublished data). While we believed that the VSV genome would be relatively protected expressed as an RNP complex, it was nevertheless possible that miRNA recognition of the genome could occur, as the masking of target sites by secondary structure in picornaviruses did not preclude miRT recognition. We therefore oriented miRTs such that they would target either the genome (g) or mRNA (r) and inserted them in previously discussed regions of the viral genome.
Viruses were rescued and titrated in the absence of sequence-complementary miRNAs, and one-step growth curves were then performed to determine the replication kinetics of the recombinant, miRNA-targeted viruses, as determined by the TCID50 (as required by the FDA) (Fig. [representative of six independent experiments]). Insertion of miRTs did not cause any slowing of virus replication on BHK cells, and all rVSVs grew to high titers in the absence of corresponding miRNAs.
miRNA-targeted VSVs are poorly responsive to sequence-complementary mimics supplemented in trans.
After constructing miRNA-targeted rVSVs, we next looked at the responsiveness of these viruses to sequence-complementary miRNAs. Viruses were analyzed to see if sequence-complementary miRNAs could provide protection from the cytolytic effects of the virus, a decrease in luciferase activity, or a decrease in viral replication.
Since the matrix protein of VSV can shut down host mRNA export, including cytokines such as IFN-β that could potentially shut down viral replication, we used cell lines that are capable of inducing and responding to cellular IFN-β (52
). A549 human lung epithelial cells were transfected with miRNA mimics corresponding to miR-125, miR-206, or a control miRNA thought to have no mammalian ortholog. Four hours after transfection of these miRNAs, A549 cells were infected with recombinant miRNA-targeted VSVs at a low MOI (0.2). While we have shown previously that picornaviruses assayed in this manner are fully protected from cytolytic effects of viral infection (26
), VSV proved to be very poorly responsive to miRNA mimics provided in trans
(Fig. ). Only the virus containing targets to miR-206 oriented to target the mRNA of the viral L gene had any significant increase in cell viability (P
= 0.0012), and this was only ~30%. Luciferase activity in these cells was not significantly different in the presence of sequence-complementary or control mimics (Fig. ).
FIG. 4. VSVs encoding neuronal targets are largely unresponsive to sequence-complementary miRNA mimics. (A) Cell viability, as determined by MTT assay, at 24 h post-viral infection at an MOI of 0.5 with recombinant miRNA-targeted VSVs pretreated for 4 h with (more ...)
In fact, the only parameter that showed a marked response was viral replication, measured by release of virus into the cell supernatant (Fig. ). Viral replication was decreased approximately 1.5 log in viruses containing miR-125T or miR-206T targeting the mRNA, with both inserted into the 3′UTR of L. While not on the same order of magnitude as that seen in picornaviruses (which can completely inhibit release of infectious virus in the supernatant), this reduction in viral titer could potentially attenuate the virus in the brain. A panel of other cell lines, including BHK, CT-26, and MEF (which represent cells with either productive or altered IFN responses), was utilized to study the miRNA responsiveness of VSVs, and these cells were equally (or less) responsive to miRNA-mediated viral inhibition (data not shown).
Neuronal target insertion protects primary brain cells from cytopathic effects in vitro.
While the expression of miRNA mimics in trans
served to effectively shut down the expression of lentiviral vectors containing miR-125T, by >95%, they had very little effect on luciferase expression in rVSVs, nor did they serve to protect cells from cytolytic effects of the virus (Fig. ). While we have shown previously that miRNA mimics are able to functionally suppress picornavirus infection in the presence of fully complementary targets (26
), we have also seen that expression of miRNAs in trans
cannot serve to functionally silence targets as efficiently as that of abundant endogenous miRNAs (Kelly et al., unpublished data). We therefore employed primary cells that express large amounts of endogenous miR-125 to see if endogenous miRNAs could indeed affect VSV replication.
VSV is capable of infecting neurons, astrocytes, and glia cells in the brain. While it is likely that neuropathogenesis is a result of cytotoxic effects on neuronal cells, cytokines such as interleukin-2 (IL-2) and nitric oxide synthase (NOS), produced by astrocytes and microglia, can have protective effects on surrounding cells (29
). To determine if the presence of miR-125 in primary astrocytes could protect these cells from infection with rVSVs bearing miR-125T in different regions of the virus, we infected primary astrocytes with miRT VSVs and looked at cytotoxicity and viral replication when cells were infected at a low MOI.
Twenty-four hours after infection with VSV (even at an MOI as low as 0.2), severe cytopathic effects were seen in primary astrocytes. Analysis by MTT assay showed that ~50% of cells were killed by rVSVs expressing control miR-206T. Insertion of miR-125T in luciferase provided no protection from cytopathic effects in primary astrocytes. However, the presence of miR-125T in the M protein rescued cell viability by roughly 15%, while miR-125r in L completely restored cell viability, such that it was not significantly different from that of mock-infected cells (P = 0.515) (Fig. ). In addition, miR-125r in L caused a >2-log reduction in viral titers released into the supernatants of infected astrocytes (Fig. ).
FIG. 5. Viral replication of miRT VSVs is restricted in primary brain cells. (A) Cell viability, as determined by MTT assay, at 24 h post-viral infection at an MOI of 0.5 with recombinant miRNA-targeted VSVs in primary astrocytes. (B) Viral titers collected at (more ...)
To validate that primary astrocytes did indeed contain miR-125, we performed TaqMan miRNA expression analysis of these cells to confirm that this brain-specific miRNA was upregulated compared to that in cells of other lineages (Fig. ). While miRNA mimics added in trans
could not significantly protect cells from VSV toxicity (Fig. ), here we show that higher levels of endogenous miRNAs expressed in astrocytes enabled VSV to become responsive to cellular miRNAs. While this result is modest compared to the effects seen in picornavirus-infected cells (5
) or adenovirus-infected cells (55
), it is nevertheless statistically significant (P
< 0.001). Decreases of viral titers in vivo
by >2 log could certainly provide protection from viral toxicities.
miRT insertion protects against neuropathogenesis when VSV is administered intracerebrally.
Having determined that rVSVs did respond to cellular miRNAs in vitro, we next proceeded with in vivo evaluation of the toxicity profiles of these viruses. While neuropathology develops from VSV infection with almost any administration route in immunocompromised rodent models, intracranial inoculation clearly provides the largest burdens for safety evaluation.
To assay for VSV attenuation in the brain, we therefore delivered 2× LD50 of VSV-Luc. This was also the LD100 for the virus (Fig. ). Immunocompromised mice were injected with 1 × 104 rVSV particles and imaged for bioluminescence 2, 5, 7, 10, 12, 14, 28, 35, and 42 days after inoculation with the virus. After virus administration, animals were monitored closely for signs of neurotoxicity, and no complications were seen in animals injected intracerebrally with control carrier.
Based upon in vitro analysis of our recombinant viruses (as well as small pilot experiments), we chose three recombinant, miRNA-targeted viruses for further evaluation (as well as the parental VSV-Luc). Viruses with miR-125T targeting the mRNA (125r) inserted in either the M or L gene were analyzed, as was a virus used as a control for insertional attenuation, containing miR-206r in L. Viral replication visualized by luciferase expression was seen at 24 h postinoculation in animals treated with VSV-Luc and the control VSV 206r L. In addition, the insertion of miR-125T in M did not attenuate virus replication in the brain 2 days after inoculation. However, only 1 of 10 animals administered 1 × 104 VSV 125r L particles had any detectable luciferase signal at the same time point (data not shown).
One week after delivery of rVSV, severe neurotoxicity had developed in groups treated with VSV-Luc, VSV 125r M, and VSV 206r L. Intense luciferase expression (>1 × 107 photons/s/brain) was seen in the majority of these animals 7 days after virus administration (Fig. ). However, 9 of 10 animals treated with VSV 125r L had no luciferase expression in the brain and no symptoms of neurotoxicity. These animals remained healthy and negative for luciferase expression up to 35 days postinoculation. At this time, all animals were euthanized and brains were harvested for tissue overlay to look for viral recovery or were sent for histological examination.
FIG. 6. VSVs encoding neuron-specific miR-125T have reduced neurotoxicity. Mice were inoculated intracranially with 1 × 104 rVSV particles and were imaged at the indicated times, using an IVIS200 system (Xenogen Corp., Alameda, CA), and monitored for (more ...)
After intracranial virus administration, 90% of animals treated with 1 × 104 VSV 125r L particles survived to the terminal point of the study (35 days) (Fig. ). This survival rate was highly significantly different from those with parental VSV-Luc (P < 0.001) and the control for insertional attenuation (P = 0.0025). While animals treated with VSV-125r M did trend toward increased survival, with 30% surviving to the terminal point in the study, this survival rate was not significantly increased over that with control miRT. Examination of luciferase expression in the brain also revealed that VSV 125r L was attenuated in the brain, with only one animal treated with the virus ever becoming positive for luciferase expression (and quickly succumbing to lethal encephalitis) (Fig. ). The observation that 10% of mice manifested with neurotoxicity even when viral expression was targeted by sequence-complementary miRNAs in the brain provides further evidence that this targeting paradigm, while effective for negative-strand viruses, is not as efficient as that with positive-strand picornaviruses and lentiviral vectors.
While the inclusion of target elements for miR-125 targeting the mRNA of the L gene of VSV did significantly attenuate the virus in the brain and increase the therapeutic index of VSV, mutation of these target elements is always of potential concern. Because only ~100 nt of sequence insert provided the entire basis of the reduced toxicity for this virus, we looked at the potential of this virus to mutate the sequence insert. In vitro selective pressure did not provide any change in the sequence insert of the virus, nor did the animal treated with VSV 125r L that succumbed to neurotoxicity (Fig. ) have any sequence alterations in the insert sequence. However, in a pilot study in which 1 × 104 VSV 125r L particles were administered to immunocompromised mice, we did see one animal become positive for luciferase expression 4 weeks after administration. The signal gained in intensity over a period of 2 weeks, finally manifesting in the spinal cord, in addition to expression in the brain (Fig. ). Forty-two days after inoculation, this animal was sacrificed, and its brain was harvested to recover virus. This virus was then amplified on BHK cells in vitro, and the miRT insert and flanking region were cloned into TOPO and sequenced, but no sequence alterations in the miRT insert were found. Therefore, the delayed development of VSV encephalitis could be a result of second-site mutations, or perhaps saturation of endogenous miR-125 as a result of VSV infection.
In vivo evaluation of recombinant miRNA-targeted VSVs showed that insertion of target elements for miR-125T targeting the L gene of the virus significantly ameliorated neurotoxicity. However, as with all viral engineering, selective pressure to mutate this sequence does occur and can hamper the efficacy of this targeting paradigm, even in negative-strand viruses.
miRNA-targeted VSVs retain oncolytic efficacy in vivo.
Our previous experiment showing that VSV incorporating miRNA targets within the viral genome can actually attenuate the virus in the brain is of great import. However, since these viruses are to be used therapeutically as anticancer agents, we next wanted to validate that they did indeed retain their oncolytic efficacy in models of oncolysis in vivo.
miRNAs are known not only to have tissue-specific signatures but to have cancer-specific expression profiles as well (12
). While there are miRNAs that are indicated as oncosuppressive and are found to be downregulated significantly in tumors, so too are there miRNAs that are oncogenic and found to be highly enriched in tumors (23
). miR-125, in fact, has actually been indicated to have a role in breast cancer. While miR-125 is most highly enriched in the brain, it is also expressed (though at much lower abundance) in normal breast tissue (24
). However, in many breast cancers, miR-125 is found to be downregulated significantly or even altogether absent.
While downregulation of miRNAs in tumors should pose no barriers for effective oncolysis by viruses that express these miRTs, increased expression of miRNAs in tumors could hamper their efficacy. While there have been no published works suggesting that any brain-specific miRNA is upregulated in cancerous cells of any origin, we nevertheless decided to test that miRNA-regulated viruses did in fact retain oncolytic efficacy.
Many groups have published that mouse colorectal cancer is particularly susceptible to VSV oncolysis (21
). Since CT-26 cells are colorectal carcinoma cells derived from BALB/c mice, they provided a good model system, since they could be implanted as subcutaneous grafts and studied in fully immunocompetent mice. Therefore, 5 × 106
CT-26 cells were injected subcutaneously into the right flanks of immunocompetent BALB/c mice. When tumors reached approximately 250 mm3
, they were treated with one intratumoral dose of 1 × 109
rVSV on day 0 and one intravenous dose of 1 × 109
rVSV on day 3. Tumors were monitored daily, and bioluminescence imaging was performed on days 2, 5, 7, and 10 after first virus administration.
Control tumors grew quickly and were unencumbered by control carrier treatment, and all animals were euthanized by day 18 after treatment (when tumors reached >2,000 mm3 or ulcerated). Animals treated with parental VSV-Luc or VSVs containing miRT inserts showed prolonged tumor progression (Fig. ) and survival (Fig. ), but tumors in all groups eventually grew such that they were larger than 10% of body weight (>2,000 mm3) or ulcerated, necessitating euthanasia of all animals due to tumor size.
FIG. 7. MicroRNA-targeted VSVs have equivalent antitumor activities in CT-26 model. (A) CT-26 cells (5 × 106) were injected subcutaneously into the right flank of BALB/c mice, and mice were administered one intratumoral dose of virus (1 × 109 (more ...)
To visualize viral replication in tumors (as well as the potential for neurotoxicity), we imaged all mice at days 2, 5, 7, and 10 after the first virus administration. At 2 days posttreatment, virus was present at similar abundances in mice treated with VSV-Luc or miRT-targeted viruses, but luciferase activity quickly decreased, such that no viral replication was detected in subcutaneous tumors or other tissues by 10 days posttreatment (Fig. ). While we did observe a luciferase signal in the head for 1/8 mice in each group (Fig. ), it never appeared to be a result of neurotoxicity. These animals had no symptomatic manifestation of neuropathogenesis, and upon sacrifice and harvest of the brain, no virus was detected. We therefore hypothesize that this may actually be a signal in the mucosa of the mouth and/or nose. There was no significant difference in antitumor efficacies of miRT recombinants (P = 0.2353), nor was there a significant difference in viral replication in tumors, as analyzed by luminescence activity on the right flank (P = 0.9294).