INTRODUCTION
Several years ago, we proposed consideration of recombinant, nonpathogenic poliovirus (PV) for the treatment of glioblastoma (GBM) (
17). This proposal is based on widespread ectopic expression of the PV receptor, nectin-like molecule-5 (Necl-5), in such cancers (
26). Necl-5, an onco-fetal cell adhesion molecule of the nectin family, is broadly associated with ectodermal/neuroectodermal cancers (reviewed in reference
42). Necl-5 expression is abundant in GBM cells, “stem cell-like” GBM cells, and tumor-associated vasculature (
6) and is implicated in GBM cell dispersion and invasion (
39,
40). Due to Necl-5 expression, GBM cells are naturally susceptible to infection with and rapid destruction by PV (
17). Direct cytocidal effects of PV elicit host immunogenic responses directed against tumors
in vivo (
43).
Any clinical application of engineered PVs must include a rigorous demonstration of safety in established nonhuman primate models for paralytic poliomyelitis. Such safety studies are modeled after standard neurovirulence assays for the live-attenuated (Sabin) PV vaccines (
41). The three Sabin vaccine serotypes (PV1-S, PV2-S, and PV3-S), some stemming from serial passage in diverse simian tissue culture systems, exhibit substantially reduced primate neurovirulence (
36). While the genetic base for attenuation is different for each Sabin strain (
29), they have key sequence variables in common.
PV plus-strand RNA genomes are not equipped with a 7-methyl-guanidine cap (
31) and thus are unable to recruit ribosomal subunits via the cap-binding eukaryotic initiation factor (eIF) 4E. Instead, PV RNAs rely on an internal ribosomal entry site (IRES) within their 5′ untranslated region (UTR) to recruit 40S ribosomal subunits internally, in a 5′ cap-, eIF4E-independent manner (
21,
34). This occurs by direct binding of the central scaffold of the translation initiation apparatus and “ribosome adaptor” eIF4G to the IRES (
8). Intriguingly, the three Sabin serotypes feature point mutations within a defined region of IRES stem-loop domain (SLD) V (
45) that may impede eIF4G binding and, hence, virulence in normal neuronal cells (
8,
33). The Sabin SLD V mutations are critical attenuation determinants, since reversion restores neurovirulence (
12,
23). Accordingly, in cases of Sabin vaccine-associated poliomyelitis, reversion of the IRES attenuating mutation to the wild-type (wt) sequence is invariably detected in the culprit virus isolated from affected patients.
Non-neurovirulence of the PV variant studied here does not rely on IRES point mutations but on a heterologous IRES derived from human rhinovirus type 2 (HRV2) (
15). Insertion of the HRV2 IRES into wt PV1, yielding PV1-RIPO, reduced neuronal competence below the levels of PV1-S, evident as poor replication potential in neuroblastoid cell lines, e.g., Sk-N-Mc (
15) or HEK-293 (
5,
14). Accordingly, PV1-RIPO lacked neurovirulence in the World Health Organization (WHO) standard neurovirulence test after intraspinal inoculation in
Macaca fascicularis (
16). Remarkably, HRV2 IRES substitution alone achieves attenuation levels similar to PV1-S (the PV1-S consensus sequence features 57 potentially attenuating mutations, 21 of which result in coding changes [
32]). Studies with PV1:HRV2 mix-and-match IRES recombinants suggest a mechanistic base for attenuation similar to the Sabin attenuating SLD V mutations, since viral competence in neuronal cells covaries with IRES portions implicated in eIF4G binding (
5,
16).
To derive a maximally attenuated PV, we inserted the HRV2 IRES into PV1-S, yielding PVS-RIPO (
10). We report here on dose-range-finding, toxicology, biodistribution, shedding, and neutralizing antibody studies of PVS-RIPO after intrathalamic challenge in
M. fascicularis.
DISCUSSION
PVS-RIPO, which is fully replication competent in susceptible tissue culture cells (e.g., Vero [
C. aethiops kidney] cells) or the intended GBM target (
17), exhibits extraordinarily low neurovirulent potential after intrathalamic inoculation in
M. fascicularis. Inoculation of up to 5 × 10e9 TCID
50 of PVS-RIPO did not produce neurohistopathological lesions, clinical neurological symptoms, or other signs of active viral replication in the primate CNS. There were no hematologic or clinical chemical parameters consistent with active viral replication. There was no evidence for viremia, extraneural spread, extraneural replication, or shedding of PVS-RIPO with saliva, urine, or stool. Our findings suggest that the intrathalamic virus inoculum remains contained within the CNS of injected animals. The primate neurovirulence assays described here are concordant with the results of previous empirical studies in neuron-like tissue culture models (
5,
16,
17,
27,
48), in mice transgenic for human
Necl-5 (
15,
16) and in
M. fascicularis challenged with intraspinal inoculation of PV1-RIPO (
16). Our investigations confirm a dominant role for the IRES and, hence, ribosome recruitment to viral RNAs in PV pathogenesis.
Mechanistic studies of neuronal incompetence of the foreign HRV2 IRES in PVS-RIPO revealed deficits in ribosome recruitment. Since PVS-RIPO translation is unencumbered in transformed cells, such deficits are likely due to tissue type-specific host conditions in the normal CNS. For the type 1 picornaviral IRES, ribosome recruitment relies on direct interactions with eIF4G (
8). Thus, the PV-S/HRV2 IRES may exhibit deficits attracting eIF4G in the normal CNS. Indeed, genetic signatures responsible for neuroattenuation mediated by the PV-S IRES (
8,
33) or the HRV2 IRES (
5,
16) map to structures that participate in eIF4G contacts.
Multiple, not mutually exclusive events may dictate IRES function in the host CNS by controlling eIF4G-IRES interactions. Most directly, PVS-RIPO translation may be determined by the functional state of eIF4G itself. In GBM cells, supporting rampant PVS-RIPO translation (implying efficient IRES function [
17]), constitutively active signal transduction pathways converge on eIF4G and its binding partners. We showed that activation of the protein kinase C-α (PKC-α)-Erk1/2 mitogen-activated protein kinase (MAPK) signaling axis with phorbol esters leads to phosphorylation of eIF4G(Ser1186) (by PKC-α) and eIF4G(Ser1232) [by a currently unknown kinase(s)] (
9). Phosphorylation of eIF4G(Ser1186) and activation of Erk1/2 MAPKs induces MAPK signal integrating kinase 1 (Mnk1) binding to eIF4G (
38), which is required for phosphorylation of eIF4E by Mnk1 (
35). We showed that signaling through Erk1/2 to MAPK signal integrating kinase 1 (Mnk1) determines PVS-RIPO translation and cytotoxicity in GBM (
14). Accordingly, providing constitutively active Mnk1 in neuron-like cells stimulates viral, HRV2 IRES-mediated translation and reverses the CNS deficit of PVS-RIPO (
14). MAPK signal transduction to protein synthesis machinery may favor viral cap-independent translation by stimulating the formation of translation-competent RNPs via recruitment of eIF4G to the IRES.
Indirectly, IRES-eIF4G interactions may be influenced through host RNA-binding, IRES
trans-acting factors (ITAFs). The double-stranded RNA-binding protein 76 (DRBP76) associates with the HRV2 IRES specifically in neuronal cells (
27) and intercepts IRES-mediated translation initiation (
28). A physiologic function of DRBP76 in translation repression at invading viral genomes, possibly as part of an antiviral defense mechanism, has been proposed (
18). Since ITAFs may also favor eIF4G-IRES binding (
22), positive contributors to an IRES ribonucleoprotein (RNP) complex conducive to eIF4G recruitment may be lacking in the CNS or may be displaced by DRBP76. We showed that DRBP76 exhibits distinct properties in GBM compared to brain. Most importantly, it is almost exclusively nuclear in GBM versus predominantly cytoplasmic in brain (
30). Thus, lacking ribosome recruitment at the HRV2 IRES in the normal CNS may reflect the array of cytoplasmic RNA-binding proteins encountered by PVS-RIPO.
For comparison of their neurovirulent potential, we contemplated including the PVS-RIPO precursor, PV1-S, in our studies. Since the studies were designed to support regulatory evaluation of PVS-RIPO, this was not pursued. Although they cannot replace primate studies, mechanistic investigations of neuronal IRES competence offer insight into this issue. Compared to wt PV1, both PV1-S and PVS-RIPO share significant translation, replication, and cell-killing deficits in HEK-293 neuroblastoid cells (
5). These are far more pronounced for PVS-RIPO compared to PV1-S (
48), suggesting lower neuronal IRES competence due to inherently impaired eIF4G recruitment to the foreign HRV2 IRES. This assumption is supported by differences in the genetic basis for neuronal IRES deficits. In PV1-S, they rely on a single nucleotide substitution, G480A, in IRES SLD V (
8,
33). In contrast, IRES incompetence of PVS-RIPO rests on a broad, discontinuous, and structurally complex base comprising SLDs V and VI in the HRV2 IRES (
5,
16). Interestingly, eIF4G/eIF4A interactions with picornavirus type 1 IRES involve the same structures (
8).
This contrast is also evident when considering the most significant safety shortcoming of PV-S, genetic instability. The G480A IRES mutation in PV1-S spontaneously and rapidly reverts to wt upon propagation in Vero cells (
19). Such adaptation events do not occur with PVS-RIPO (
10). This suggests that eIF4G recruitment deficits of PV1-S, due to a single IRES mutation acquired upon artificial selection
in vitro, are shallow compared to those in PVS-RIPO, encrypted by a naturally evolved IRES element. Therefore, we stipulate that neuroattenuation of PVS-RIPO is at least at level with PV1-S, if not significantly lower. This is supported by neurovirulence tests of PV1-S and PV1-RIPO after intraspinal inoculation in
M. fascicularis (
16).
We document safety of PVS-RIPO in terms of low neuropathogenicity, inability to disseminate or replicate extraneurally, and lack of shedding. Our studies support the clinical use of PVS-RIPO as a direct oncolytic, indirect immunotherapeutic agent against GBM.
ACKNOWLEDGMENTS
This project has been funded in whole or in part with federal funds from the National Cancer Institute, National Institutes of Health, under contract NO1-CO-12400 and also supported in part through the NCI-RAID Program of the Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis, National Cancer Institute, National Institutes of Health.
This study was partly supported by PHS grants CA124756 and CA140510 (M.G.). The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.
We thank A. Nomoto for providing the PV1-S cDNA clone and J. H. Sampson and his team for sharing qRT-PCR equipment.