hsp72, a Host Determinant of Measles Virus Neurovirulence

Abstract

Transient hyperthermia such as that experienced during febrile episodes increases expression of the major inducible 70-kDa heat shock protein (hsp72). Despite the relevance of febrile episodes to viral pathogenesis and the multiple in vitro roles of heat shock proteins in viral replication and gene expression, the in vivo significance of virus-heat shock protein interactions is unknown. The present work determined the in vivo relationship between hsp72 levels and neurovirulence of an hsp72-responsive virus using the mouse model of measles virus (MV) encephalitis. Transgenic C57BL/6 mice were created to constitutively overexpress hsp72 in neurons, and these mice were inoculated intracranially with Edmonston MV (Ed MV) at 42 h of age. The mean viral RNA burden in brain was approximately 2 orders of magnitude higher in transgenic animals than in nontransgenic animals 2 to 4 weeks postinfection, and this increased burden was associated with a fivefold increase in mortality. Mice were also challenged with an Ed MV variant exhibiting an attenuated in vitro response to hsp72-dependent stimulation of viral transcription (Ed N-522D). This virus exhibited an attenuated neuropathogenicity in transgenic mice, where mortality and viral RNA burdens were not significantly different from nontransgenic mice infected with either Ed N-522D or parent Ed MV. Collectively, these results indicate that hsp72 levels can serve as a host determinant of viral neurovirulence in C57BL/6 mice, reflecting the direct influence of hsp72 on viral gene expression.

Cellular heat shock proteins (HSPs) are recognized for their function as molecular chaperones that facilitate protein folding and trafficking (22) and for their ability to bind native proteins, resulting in altered activity of the substrate (18). Multiple families of HSPs are recognized and are classified according to their mass, with members of the 70-kDa family being expressed at high basal and/or stress-inducible levels in multiple tissues. In particular, the major inducible 70-kDa HSP (i.e., hsp72, also known as hsp70) exhibits a wide range of expression levels in the cytosol, being readily induced by physiological stimuli, such as fever (32, 44). This dynamic range of expression and numerous roles in protein metabolism make hsp72 a potentially significant variable that could influence the outcome of viral replication in an animal host. In vitro studies suggest that hsp72 and the constitutively expressed isoform can directly modulate gene expression of several mammalian RNA and DNA viruses by supporting viral core protein maturation and/or assembly into nucleocapsid particles (11, 12, 30, 31), by mediating assembly and/or activity of viral polymerase/replication complexes (7, 29, 45), or by mediating nuclear trafficking of viral preintegration complexes in the case of retroviruses (1). Despite the relevance of hsp72 to viral replication within the cell and the physiological relevance of elevated hsp72 to viral infection of an animal host, the biological (in vivo) significance of virus-hsp72 interactions is unknown.

The objective of the present work was to establish the in vivo significance of virus-hsp72 interactions using a viral system in which the basis for hsp72-dependent changes in viral gene expression is defined and therefore can be manipulated. Elevated levels of hsp72 increase transcription and genome replication of Edmonston measles virus (Ed MV), leading to increases in cytopathic effect (CPE) and/or cell-free progeny release in multiple cell lines (9, 40, 48, 49, 53). The mechanism involves, at least in part, binding of hsp72 to the nucleocapsid template for the viral RNA-dependent RNA polymerase. Specifically, hsp72 recognizes two conserved hydrophobic domains on the exposed C terminus of the nucleocapsid protein (N) (i.e., Box-2 and Box-3) that are also recognized by the viral polymerase cofactor P (nucleocapsid-associated phosphoprotein) (5, 25, 53, 54). The P protein serves as a tether between the nucleocapsid template and the viral polymerase, and the high binding affinity between P and N indicates the need for a cofactor that would loosen the complex in order to promote cycles of binding and release that would be required for polymerase processivity (4, 5). hsp72 is thus a prime candidate for such a cofactor, having been shown to directly compete with the P protein (i.e., the X domain) for Box-2 binding (53). hsp72 may also destabilize P-N complexes by binding Box-3. Previous work has shown that a single amino acid substitution (N522D) in Box-3 of Ed MV can selectively disrupt hsp72 binding (54). Virus incorporating this variant motif (Ed N-522D) exhibits a significantly attenuated transcriptional response to elevated hsp72 levels, whereas hsp72-dependent increases in genome levels are identical between Ed and Ed N-522D virus (9, 53). Basal profiles of gene expression and replication are otherwise identical for parent and Ed N-522D viruses, making Ed MV and the Ed N-522D variant ideal tools to dissect the role of functional hsp72-virus interactions in altering viral virulence.

The present work examines the effect of elevated hsp72 on Ed MV virulence using a mouse model of MV encephalitis. The hsp72 responsiveness of Ed MV is well defined in mouse neuroblastoma cells (9), and neurons are a primary target of infection in mouse models of Ed MV encephalitis (3, 34). Challenging neonatal C57BL/6 mice via the intracranial route exploits both the mouse age- and strain-dependent susceptibilities to MV infection (16, 20, 27, 34, 36), thereby obviating the need for transgenic expression of human membrane cofactor protein (CD46) for use as an ancillary viral receptor or use of mouse brain-adapted strains of MV in order to promote brain infection (13, 26, 33, 41). Moreover, the high incidence of stable persistent infection that results from the infection of neonatal C57BL/6 mice allows us to detect hsp72-mediated changes that either promote or diminish virulence. Mice lack a febrile response following intracranial viral inoculation, and they lack the low constitutive levels of hsp72 that can be observed in other animal species (8, 39). Accordingly, selective transgenic overexpression of hsp72 in neurons can be used to assess the influence of hsp72 on viral virulence when compared to the outcome of infection in mice of the parental strain.

Neuronal infection is relevant to MV infection of its natural primate hosts. MV is disseminated to multiple tissue compartments of humans following initial replication in lymphoid tissues. Clinical signs of disease reflect respiratory tract involvement with secondary bacterial infection. Measles virus invasion of the central nervous system (CNS) is common, albeit clinically silent in the majority of cases (6, 24). CNS complications of MV infection include measles inclusion body encephalitis (MIBE), an acute fulminate infection of neurons by either wild-type or vaccine virus in immunocompromised patients, or subacute sclerosing panencephalitis, a manifestation of viral persistence in neurons and glia that occurs years following early childhood exposure to MV (14). Aside from the differences in immune function, host variables determining the outcome of MV infection of the CNS are unknown. Our results will show that increased hsp72 levels can be a host determinant of neurovirulence for virus that is responsive to hsp72-dependent enhancement of transcription.

MATERIALS AND METHODS

Generation of transgenic (TG^+/+) mice.

The human hsp70A cDNA was placed under control of the rat neuron-specific enolase (NSE) promoter. The NSE promoter sequence was PCR amplified from the pNSE-TAK vector with Pwo DNA polymerase using the sense (5′-GTGGTACCTATGG TGGTATGGCTGAC-3′) and antisense (5′-GTAGATCTGGACACGAGACATTGGG TA-3′) primers that contain KpnI and BglII restriction sites, respectively (underlined). The 1,789-bp amplicon was inserted into the pBluescript L7 expression cassette. Transcripts produced from this cassette exhibit increased stability in neurons (37, 43). A 2,135-bp fragment containing the human hsp70A cDNA was excised from pT7VSV-hsp72 using KpnI and BamHI and inserted downstream of NSE into a unique BamHI site within the L7 cassette. Inducibility of the resultant construct within its plasmid vector (pNSE-VSV/hsp72) was tested in differentiated mouse neuroblastoma cells (ATCC CCL-131, also called N2a) following calcium phosphate transfection (54). Differentiation was induced by treatment with 20 mM LiCl and 10 mM myo-d-inositol as previously described (19). Neuronal differentiation was based upon morphological characteristics and expression of NSE protein. Whole-cell protein lysates were analyzed by Western blotting using primary antibodies specific for NSE (AB951; Chemicon), hsp72 (SPA-810; StressGen), or vesicular stomatitis virus G protein (VSV-G) (P5D4; Roche).

The construct was linearized with KpnI and XbaI and gel purified, and the 5.9-kb fragment was microinjected into fertilized oocytes from female C57BL/6 mice for subsequent implantation. Transgenic founders were identified by PCR of genomic DNA isolated from tail biopsy specimens. A 526-bp amplicon spanning L7 and hsp70A was amplified using 5′-TTGGAGGCACTTCTGACTTGCA-3′ and 5′-TTCGCGTCAAACACGGTGTT-3′ sense and antisense primers, respectively. An L7-hsp72-specific cDNA probe was generated by random primer extension using DNA polymerase I, [α-³²P]dCTP, and the 526-bp PCR amplicon generated from pNSE-VSV/hsp72 for Southern blot analysis of genomic RNA. Transgene copy number was determined by phosphorimage analysis comparing the intensities of transgene bands to those from serial dilutions of XbaI-linearized plasmid (pNSE-VSV/hsp72).

Total RNA was isolated from tissues for reverse transcriptase PCR (RT-PCR) analysis of transgene expression. Tissues were homogenized in Qiazol lysis reagent (QIAGEN), and total RNA from brain was isolated using the RNeasy lipid tissue mini kit (QIAGEN) and from nonneural tissues using the RNeasy mini kit (QIAGEN). Viral cDNA was synthesized with Moloney murine leukemia virus reverse transcriptase (Invitrogen) using the sense primer 5′-ATGGCTTACACCGACATCGAGA-3′ that sits on the VSV-G epitope, thereby distinguishing exogenous from endogenous hsp72. PCR was performed using the sense and the antisense primers described above. Tissues were also homogenized in lysis buffer (20 mM HEPES, 1% Triton X-100, 30 mM sodium pyrophosphate, 50 mM NaF, 1 mM EGTA, 2 mM EDTA, 1 mM Na₃VO₄, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, and 1 μg/ml aprotinin), and total protein concentration was measured by the Bradford method (Bio-Rad). Transgenic hsp72 protein (VSV/hsp72) was detected by Western blot analysis using a mouse monoclonal antibody recognizing the VSV-G epitope tag (P5D4; Roche), a horseradish peroxidase-conjugated goat anti-mouse secondary antibody (Chemicon), with the signal detected by chemiluminescence. Localization of transgene expression in brain was based upon indirect immunohistochemical staining of the VSV-G epitope in 5-μm formalin-fixed paraffin-embedded tissue sections. The primary antibody was a rabbit polyclonal antiserum, used at a dilution of 1:800 (PRB-192P; Covance Research Products). Secondary anti-rabbit antibody was labeled with horseradish peroxidase, and signal was developed with 3-amino-9-ethylcarbazol substrate.

Characterization of infected mice.

Male and female TG^+/+ and nontransgenic (NT) neonatal mice (42 h old) were inoculated intracranially in the left cerebral hemisphere with Ed MV or Ed N-522D MV in a total volume of 10 μl as previously described (8). Control mice received either tissue culture media or an equivalent dose of virus that was irradiated with 6 × 10⁴ μJ of short-wavelength UV light. Mice were euthanized, and the brains were harvested at 21 days postinfection (PI), 70 days PI, or when animals met early removal criteria (i.e., weight loss of >30 to 50% of the mean litter body weight, dehydration, and seizure activity). Brains were sagittally sectioned. The right half was processed for total RNA isolation, and the left half was processed for either light microscopic analysis of tissue sections (frozen or formalin-fixed) or titration of viral progeny by cocultivation of fresh brain homogenates on Vero cells (51).

Levels of viral RNA were quantified by SYBR green real-time RT-PCR as previously described (8). The one-step RT-PCR targeted both plus- and minus-sense N-gene sequence (i.e., viral transcript and antigenome as well as genome RNA), thereby maximizing sensitivity. Formalin-fixed paraffin-embedded tissue sections were evaluated for CPE following routine hematoxylin and eosin staining. In situ hybridization of viral N-gene sequences was performed on 5-μm formalin-fixed sections as previously described (10). Uninfected and sham-infected tissue sections were run as controls. The MV N-specific probe was amplified by PCR using pT7MV-N as a template and sense and antisense primers corresponding to genome nucleotide positions 1367 to 1388 and 1589 to 1609, respectively. The 234-bp PCR amplicon was labeled by random priming with dioxigenin-dUTP. Frozen sections were acetone fixed and used for immunohistochemical staining of viral hemagglutinin (H) glycoprotein. Three mouse monoclonal antibodies recognizing the MV H protein (i.e., L77, K17, and K77) were used in a mixture of 1:1:1 at a dilution of 1:100 (16, 46). Sections from uninfected and sham-infected mice were included as controls. In addition, naïve mouse immunoglobulin G was used as an antibody-specific isotype control for the primary antibody.

RESULTS

Generation and characterization of transgenic mice.

An expression construct was designed for expression of hsp72 bearing a VSV-G epitope tag under control of a neuron-specific enolase promoter (see Fig. 2A). Expression of the construct within its plasmid vector (pNSE-VSV/hsp72) was first tested in vitro following transfection of both undifferentiated and differentiated mouse neuroblastoma (N2a) cells. LiCl induced maximal morphological differentiation in N2a cells by 48 h, at which time both NSE and VSV/hsp72 was detected in differentiated cells, but not undifferentiated cells, by Western blot analysis of total cell protein (Fig. 1).

FIG. 2.

Expression of the VSV-G-tagged hsp72 in TG mice. (A) The linear 5,957-base-pair NSE-VSV/hsp72 expression construct generated by KpnI/XbaI digestion of pNSE-VSV/hsp72. Arrows indicate the primer positions used for RT-PCR analysis of construct expression in the tissues of mice (see below). The upstream primer includes sequence of the VSV-G tag, thereby distinguishing between transgenic and endogenous hsp72. Both the VSV-G tag and hsp72-coding sequence are placed within the L7 cassette, which stabilizes resultant transcripts in neurons. (B) RT-PCR analysis of transgene expression from the F₁ progeny of founder 5769. Negative controls included reaction mixtures with no template (−) and reaction mixtures lacking RT. Total RNA was isolated from neural and nonneural tissue. The positive control used pT7VSV-hsp72 as a template. The 266-bp amplicon (arrow in 5769 gel) reveals tissue-specific expression from the NSE promoter. Western blot analysis of total protein extracts was run in parallel with the RNA analysis. Antibody specific for the VSV-G epitope distinguished transgenic hsp72 (arrow) from endogenous hsp72. The negative control was total protein from the cerebellum of a NT mouse (−), and the positive control was total protein from Vero cells transfected with pT7VSV-hsp72. G3PDH, glyceraldehyde-3-phosphate dehydrogenase. (C) Homozygotes (TG^+/+) were generated from founder 5769. Immunohistochemical staining of the VSV-G epitope in sections of cerebral cortex (top panels) and hippocampus (bottom panels) revealed neuronal expression of VSV/hsp72 in TG^+/+ mice (right panels) but not in NT mice of the parental strain (left panels).

FIG. 1.

Expression of NSE-VSV/hsp72 in differentiated cells but not in undifferentiated murine neuroblastoma cells. Cells were transfected with pNSE-VSV/hsp72 (+), and expression of the construct was detected by Western blot analysis of total cell protein using antibodies specific for either the VSV-G epitope tag or hsp72 (α VSV-G or α hsp, respectively). Expression was detected only in cells following LiCl-induced neuronal differentiation. The positive control was total protein from cells transfected with pT7VSV-hsp72, where T7 polymerase was provided by the recombinant vaccinia virus MVAT7 (54). The positions of molecular mass markers (in kilodaltons) are shown to the right of the blots.

One mouse line was established [C57BL/6-TgNSE(VSV/hsp72)5769OGL, abbreviated 5769] that supported high levels of hsp72 expression from the construct on the basis of Western blot analysis of tissue total protein extracts. The VSV-G signal was detected in the frontal cortex, hippocampus, and cerebellum but not in the liver, kidney, spleen, or myocardium (Fig. 2B). Endogenous hsp72 was not detectable when hsp72-specific antibody was used to probe protein extracts from nontransgenic littermates, and the level of the constitutively expressed isoform of 70-kDa heat shock protein (i.e., hsp73) was not affected by enforced hsp72 expression (data not shown). The latter parallels previous findings where hsp72 overexpression driven by the beta actin promoter has no effect upon the in vitro levels of hsp73 in murine neuroblastoma cells or the in vivo levels of hsp73 in the brains of transgenic mice (9, 21). The tissue-specific expression pattern was confirmed by RT-PCR analysis of total RNA isolated from both neural and nonneural tissue of transgenic mice and NT littermates (Fig. 2B). Southern blot analysis of total DNA isolated from tail biopsy specimens showed that line 5769 contained approximately two copies of the transgene (data not shown). This line was subsequently maintained as homozygotes for the transgene (TG^+/+). Expression of the recombinant hsp72 was localized to neurons of TG^+/+ mice on the basis of immunohistochemical staining of VSV-G in sections of cerebrum and hippocampus, consistent with the cell lineage specificity of NSE promoter activity (Fig. 2C).

Ed MV infection of TG^+/+ and NT mice.

Neonatal NT (n = 53) and TG^+/+ (n = 41) mice were inoculated intracranially with 4 × 10⁴ 50% tissue culture infective doses (TCID₅₀) of Ed MV. Animals were monitored daily over the 70-day course of the study. Mortality was observed between 10 and 28 days postinfection and was preceded by a 30 to 50% reduction in body weight relative to the mean weight of littermates and hyperexcitability, leading to frequent seizures, progressive weakness, and paralysis. This clinical progression closely paralleled that described for NSE-CD46 transgenic neonatal mice challenged with Ed MV (41). Kaplan-Meier survival curves, censored for sampling of healthy mice during the acute phase of infection, revealed that Ed MV-challenged TG^+/+ mice exhibited a fivefold increase in mortality (69.5%) relative to NT mice (15.1%) (Fig. 3). Median survival time was reduced from 70+ days in NT mice to 17 days in TG^+/+ mice (P < 0.001, log rank test for equality in survival distributions). Control mice receiving either tissue culture media or UV-inactivated virus exhibited no neurological abnormalities or mortality (data not shown).

FIG. 3.

Kaplan-Meier statistics were used to generate the survival curves for Ed MV-infected NT mice (n = 55) and TG^+/+ mice (n = 44). Curves were corrected for censoring (i.e., removal of healthy mice during the acute phase of infection for tissue analysis). The Wilcoxon and log rank test of survival both indicated that survival was significantly decreased in TG^+/+ mice relative to NT mice (P < 0.001). Cox proportional hazard modeling validated the results. This model assumes that the hazard ratio stays the same between the groups over time and that censoring is random and independent of the times when the animals die.

Total RNA was isolated from brain during the acute phase of infection, representing all cases of virus-induced mortality and healthy mice sampled 21 days PI. Real-time RT-PCR analysis of viral N-gene sequence was then used to calculate brain viral RNA burden. The increased incidence of mortality in Ed MV-infected TG^+/+ mice was associated with an approximately 100-fold increase in the mean viral RNA burden for the group during the acute phase of infection (Fig. 4A). The mean viral RNA burden was (3.5 ± 6.6) × 10⁶ copies for NT animals (n = 19) and (1.3 ± 1.1) × 10⁸ copies for TG^+/+ animals (n = 25), a statistically significant difference (P < 0.05, t test). Within treatment groups, there was a strong correlation between viral RNA burden and mortality. Data from all TG^+/+ animals were derived from individuals that were moribund. Data from the NT group included all moribund animals (n = 7) and an additional 12 animals that were healthy at 21 days PI. Moribund NT animals had the highest viral RNA burdens within the group. When the comparison of viral RNA burdens of TG^+/+ and NT mice was restricted to the moribund subset of animals, the difference in the means remained greater than 10-fold, a statistically significant difference (P < 0.001). The mean viral RNA burden for the moribund subset of NT-infected animals was (9.8 ± 9.6) × 10⁶ (n = 7). The viral burdens at 70 days PI ranged from 1 × 10³ to 1 × 10⁵ copies (in 250 ng of total brain RNA) with no statistical difference in these values between NT and TG^+/+ mice.

FIG. 4.

Viral RNA burden in total brain RNA from NT and TG^+/+ mice challenged with Ed MV. Tissues were harvested during the acute phase of infection (i.e., ≤28 days PI). (A) Mean viral RNA copy number in 250 ng total brain RNA based upon real-time RT-PCR of viral N-gene RNA. In vitro-transcribed N-gene RNAs were used as standards. The sensitivity of the assay was ≥100 copies. All samples were run in duplicate. Closed circles represent values derived from mice that became moribund during the acute phase of infection, and open circles represent values from mice that were clinically normal 21 days PI. Horizontal solid bars represent the mean viral RNA burdens for animals that were moribund, whereas the dashed line represents the mean for the entire group of both moribund and healthy animals. Signal was not detected in RNA from the brains harvested from mice inoculated with tissue culture supernatant or UV-inactivated virus (not shown). Differences in the mean burdens of NT and TG^+/+ mice were statistically significant (P < 0.05, t test). (B) Northern blot analysis of viral RNA was performed on samples from TG^+/+ and NT mice that were representative of the mean viral burden for the group (above). The N and H transcripts were detected with plus-strand-specific ³²P-labeled riboprobes. The leftmost lane of the NT group contains a sample from an animal that was clinically normal 21 days PI, whereas the other two samples were from animals that were moribund. Quantification of glyceraldehyde-3-phosphate dehydrogenase (GAPDH)-corrected signal intensities showed that differences in transcript levels were statistically significant for both N and H (P < 0.05, t test).

The increased MV RNA burden in TG^+/+ mice during the acute phase of infection was correlated to increased levels of viral transcripts based upon Northern blot analysis of total brain RNA (Fig. 4B). Transcripts analyzed represented genes that were either proximal or distal to the viral genomic promoter. Three samples from each treatment group were analyzed, being representative of the mean viral RNA burden for that group (i.e., 0.5 × 10⁶ to 6 × 10⁶ copies per 250 ng total brain RNA for NT mice and 1.0 × 10⁸ to 1.5 × 10⁸ copies for TG^+/+ mice). Samples from the NT mice represented two moribund animals and one healthy animal sampled at 21 days PI. Glyceraldehyde-3-phosphate dehydrogenase-corrected signal intensities for N and H transcripts were 5.3 and 1.5 times greater in TG^+/+ mice than in NT mice, respectively. These differences were statistically significant (P < 0.05, t test). Negative-sense genomic RNA was not detectable in these analyses.

Histological analysis of tissues harvested during the acute phase of infection revealed a relatively noninflammatory encephalopathy in both Ed MV-infected NT and TG^+/+ mice. The cytopathic effect was proportionate to the viral RNA burden, being greatest in the cerebral cortex and hippocampus, with decreasing severity of change in the brain stem progressing from the diencephalon to the myelencephalon. Cerebellular involvement was conspicuously absent. The most florid changes were associated with viral RNA burdens of ≥1 × 10⁸ copies per 250 ng total RNA and were restricted to TG^+/+ mice infected with Ed MV. Changes included neuronal intracytoplasmic inclusion body formation, neuronal syncytium formation in the hippocampus, selective neuronal necrosis, and rare intranuclear inclusion body formation (Fig. 5). Neurons were identified on the basis of their unique morphology, and the progression from normal to degenerate to necrotic cells was used to identify the latter as being of neuronal origin. The progression is illustrated in Fig. 5A. The degenerative sequence was characterized by central chromatolysis, cytosolic hypereosinophilia, nuclear vesiculation within syncytia, nuclear margination, and karyopyknosis or karyorhexis. In situ hybridization of viral N-gene sequences illustrated the predilection for viral growth in the CA1 to CA3 layers of the hippocampus and cerebrocortical neurons, and immunohistochemistry demonstrated H-glycoprotein expression that exhibited a similar distribution (Fig. 6). Neuronal syncytia were strongly positive for both viral RNA and MV H glycoprotein. Syncytia and immunohistochemical staining for H were not observed in tissue sections having viral RNA burdens that were <1 × 10⁸ copies per 250 ng RNA, and cytoplasmic inclusions and neuronal necrosis were not observed in tissues having viral RNA burdens of <5 × 10⁶ copies per 250 ng RNA. Inflammatory changes were only sporadically observed in the individual infected animals, and this was characterized by mild perivascular and leptomeningeal lymphocytic infiltrates and formation of glial nodules within the parenchyma. These inflammatory changes were equal in magnitude in NT and TG^+/+ mice.

FIG. 5.

Cytopathic effect during the acute phase of infection with Ed MV, illustrated in formalin-fixed sections stained with hematoxylin and eosin. (A) Viral eosinophilic intracytoplasmic inclusion bodies (arrows) and selective necrosis (N) were observed in cortical neurons of mice having mean viral RNA burdens that were ≥5 × 10⁶ copies/250 ng total RNA. Accordingly, these changes were observed in all TG^+/+ mice and 32% of NT mice. Virus-induced neuronal syncytia with nuclear swelling (B), formation of viral eosinophilic intranuclear inclusion bodies in neurons (arrows) (C), and marked neuronal necrosis (D) in the hippocampus were changes observed in mice having mean viral RNA burdens that were ≥1 × 10⁸ copies/250 ng total RNA. These changes were unique to Ed MV-infected TG^+/+ mice.

FIG. 6.

Distribution of viral RNA and H-glycoprotein expression during the acute phase of Ed MV infection, where the brain viral RNA burden is ≥1 × 10⁸ copies per 250 ng total brain RNA. (A) In situ hybridization of MV N-gene sequence in formalin-fixed tissue sections shows prominent signal localized to neurons in the CA1 to CA3 layers of the hippocampus, including neuronal syncytia. The inset is a higher magnification of a region of hippocampus, illustrating signal localized to syncytia. Signal is also disseminated among cerebral cortical neurons (upper right). (B) Immunohistochemical staining of MV H-glycoprotein expression in acetone-fixed frozen sections illustrates a staining pattern similar to that of the viral RNA. Prominent expression is observed in neuronal syncytia within the hippocampus, a portion of which has been magnified (inset). In situ hybridization and immunohistochemical staining did not result in signal in brain sections from sham-inoculated mice (not shown).

Expressed as a function of treatment group, mild CPE was detected in approximately 30% of Ed MV-infected NT mice during the acute phase of infection, being restricted to very sporadic selective neuronal necrosis and formation of intracytoplasmic inclusion bodies and lack of MV H-glycoprotein expression. All Ed MV-infected TG^+/+ mice exhibited CPE that ranged from mild to marked, with 60% exhibiting MV H-glycoprotein expression and syncytium formation in the hippocampus. Cytopathic effects and inflammatory changes were not observed in any tissues harvested 70 days PI (data not shown), consistent with the low brain viral RNA burden.

Ed N-522D MV infection of NT and TG^+/+ mice.

Mice were infected with Ed N-522D to determine whether hsp72-dependent increases in MV neurovirulence reflect functional interactions between hsp72 and Box-3 of MV N protein. Intracranial inoculation of neonatal TG^+/+ mice (n = 47) with 4 × 10⁴ TCID₅₀ of Ed N-522D induced an incidence of acute mortality that was not statistically different from that observed for NT mice infected with either Ed MV or Ed N-522D (Fig. 7). Mortality in Ed N-522D-infected NT mice was 22.2% versus 23.4% in TG^+/+ mice and was significantly reduced relative to TG^+/+ mice challenged with Ed MV (P < 0.001, log rank test).

FIG. 7.

Survival curves for NT (n = 36) and TG^+/+ (n = 47) mice challenged intracranially with Ed N-522D MV, an amino acid substitution variant of Ed MV with an attenuated in vitro transcriptional response to elevated hsp72 levels. Kaplan-Meier statistics were used to generate the survival curves and were corrected for censoring. The Wilcoxon test as well as the log-rank test of survival indicated that survival was not significantly different in the infected NT and TG^+/+ mice.

The mean brain viral RNA burden of Ed N-522D virus-infected animals was also unaffected by hsp72 overexpression. The mean viral RNA burdens (per 250 ng total RNA) during the acute phase of infection were (1.5 ± 2.7) × 10⁷ copies for NT mice (n = 21) and (1.0 ± 1.3) × 10⁷ copies for TG^+/+ mice (n = 24) when all animals within each treatment group were considered (i.e., moribund animals and those which were clinically normal 21 days PI) (Fig. 8). These mean burdens were not statistically significantly different from each another or from those observed in Ed MV-infected NT mice [i.e., (3.5 ± 6.6) × 10⁶ copies]. As with infection by Ed MV, there was a correlation between brain viral RNA burden and mortality. The mean viral RNA burdens within the subset of moribund animals were (3.5 ± 3.7) × 10⁷ and (2.0 ± 1.4) × 10⁷ for NT and TG^+/+ mice, respectively. These two means were not statistically significantly different. Histological features of infection during the acute and chronic phases were indistinguishable between challenge groups, exhibiting the relationship to brain viral RNA burden reported for Ed MV-infected mice. Thirty-three percent of Ed N-522D virus-infected NT mice and 46% of Ed N-522D virus-infected TG^+/+ mice exhibited mild encephalopathy during the acute phase of infection.

FIG. 8.

Viral RNA burden in total brain RNA of NT and TG^+/+ mice challenged with Ed N-522D. Tissues were harvested during the acute phase of infection (i.e., ≤28 days PI). (A) Mean viral RNA copy number in 250 ng total brain RNA based upon real-time RT-PCR of viral N-gene RNA. In vitro-transcribed N-gene RNAs were used as standards. The sensitivity of the assay was ≥100 copies. All samples were run in duplicate. Closed circles represent values derived from mice that became moribund during the acute phase of infection, and open circles represent values from mice that were clinically normal 21 days PI. Horizontal solid bars represent the mean viral RNA burdens for animals that were moribund, whereas the dashed line represents the mean for the entire group of both moribund and healthy animals. Signal was not detected in brain RNA harvested from mice inoculated with tissue culture supernatant or UV-inactivated virus (not shown). Differences in mean burdens between NT and TG^+/+ mice were not statistically significant.

DISCUSSION

The sustained elevation of hsp72 levels in the TG^+/+ mice is biologically relevant to MV infection of the human brain. Constitutive expression of hsp72 is observed in humans, and expression is also inducible (32). This situation is contrasted to the expression profile in rodents, where significant hsp72 expression is detectable only following stress response induction (8, 39). For both humans and rodents, hsp72 is localized primarily to gray matter in association with neuronal somata (47). A direct comparison of hsp72 levels in human brain versus rodent brain shows that the levels in nonstressed humans are more than 10-fold greater than the highest levels observed in heat-shocked rats or mice (39). The human tissues analyzed were from subjects without dementia who died suddenly and unexpectedly without evidence of agonal stress or fever and without evidence of neuropathology on postmortem examination. Further elevations of hsp72 above baseline levels are also likely in the MV-infected human brain, since fever is a consistent feature of naturally occurring MV infection (23) manifested prior to or concurrent with MV entry into the CNS. Febrile temperatures as low as 39°C are potent inducers of hsp72 in human brain (32), and these elevations should persist for days following induction on the basis of the observed stability of heat shock-induced hsp72 levels in rodent brain (39). The degree to which hsp72 expression in TG^+/+ mice of the present study approximates basal versus febrile levels in humans has not been established, although any increase in hsp72 in a rodent system should be viewed as a change more closely approximating the human condition. Fortuitously, it is the very lack of basal hsp72 expression in murine cells that allowed us to determine the role of this stress protein in MV neurovirulence.

Our results demonstrate that elevated expression of hsp72 in neurons, a primary target of MV infection, is sufficient to promote neurovirulence in mouse brain. Infection of nontransgenic neonatal C57BL/6 animals with Ed MV has been reported to cause variable mortality rates ranging from 0 to 30% (41, 42). Results of the present work are consistent with this low incidence of mortality, being 15 or 22% for NT mice infected with Ed MV or the N-522D variant, respectively. The hsp72-dependent increase in mortality can be attributed to increased virus-induced neuronal CPE, reflecting both quantitative (i.e., incidence of virus-induced cytoplasmic inclusion bodies and neuronal necrosis) and qualitative changes (i.e., formation of neuronal syncytia and intranuclear inclusion bodies in the hippocampus). Neuronal syncytium formation has otherwise been reported only for Ed MV infection of highly susceptible CBA mice (34). Distribution of viral CPE among tissue compartments was not affected by hsp72 overexpression, and in situ hybridization results confirmed the direct association between viral localization and CPE, providing no evidence of indirect mechanisms of CPE such as that mediated by excitotoxicity (2). Distribution of virus and viral CPE is therefore identical to that previously described for MV infection of nontransgenic and CD46 transgenic BALB/c × (C57BL/6xSJL) mice (3).

Northern blot analyses showed that changes in viral transcript levels underlie, at least in part, the differences in the overall viral RNA burdens that were observed in TG^+/+ and NT mice. Increased production of H transcripts is particularly significant, given the role of H protein in mediating both neurovirulence in rodent models (28) and syncytium formation (52). Syncytium formation restricted to the hippocampus can be explained by the close proximity of neuronal cell bodies in that location, facilitating direct cell-to-cell fusion that is otherwise restricted by levels of H production (50).

Proof that hsp72-dependent increases in viral transcription were the basis for increased CPE and mortality came from results of mice infected with Ed N-522D MV. Neurovirulence of Ed N-522D virus was unaffected by neuronal hsp72 overexpression, with acute-phase brain viral RNA burden, viral CPE, and virus-induced mortality being similar to those of NT mice infected with Ed N-522D and NT mice infected with parent Ed MV. This behavior parallels that observed in vitro for infected murine neuroblastoma cells (9). Viral RNA expression and growth kinetics are identical for Ed MV and Ed N-522D MV in control cells. Elevated hsp72 levels in stably transfected cells increase both viral genome and transcript levels for Ed MV, whereas the increased transcript levels are markedly attenuated for Ed N-522D MV, such that significant hsp72-dependent increases in CPE and formation of infectious progeny are observed only for Ed MV. In the present study, infectious virus was recovered from mice during the acute phase of infection when brain viral RNA burden was >1 × 10⁶ copies per 250 ng total RNA, but the limited number of animals for which such data was available did not permit statistical analysis of differences in infectious virus yields of challenge groups (unpublished observation).

The hsp72-dependent neurovirulence of Ed MV in C57BL/6 mice parallels virological and pathological findings in MIBE of humans. Neuronal necrosis, neuronal intranuclear and intracytoplasmic viral inclusion body formation, formation of viral antigen-positive multinucleated giant cells, recovery of infectious virus, and the paucity of a significant inflammatory response have all been reported for subjects with MIBE (14, 15, 17). Mouse strain-dependent differences in cell-mediated immune responses to MV infection may determine the clinical manifestations of hsp72-dependent increases in MV gene expression. Mouse susceptibility to MV encephalitis is determined by the H-2 haplotype that, in turn, dictates the effectiveness of MV-specific cell-mediated immune clearance from the brain (35, 36, 51). C57BL/6 mice carry the H-2^b allele and are therefore susceptible to MV brain infection. The possibility exists that the outcome of MV-hsp72 interaction may differ, depending upon the mouse strain and thus the effectiveness of adaptive antiviral immunity. Previous work from our laboratory supports this possibility (8). BALB/c mice carry the H-2^d allele and are capable of mounting efficient cell-mediated immune responses against MV. hsp72 was elevated in the brains of neonatal BALB/c mice using transient hyperthermia, at which point the mice were inoculated intracranially with 4 × 10⁴ TCID₅₀ Ed MV. The result was a significant increase in viral clearance from the brain relative to control mice, where clearance was temporally correlated to the onset of virus-specific cell-mediated immune responses.

We can use this and previous data to construct a more comprehensive model of MV-hsp72 interaction in brain. We propose that in an immunocompetent host, hsp72-dependent increases in MV gene expression can facilitate immune clearance by overcoming host-restricted low levels of viral gene expression that otherwise promote viral persistence (8, 38). In this context, host immune responses are adequate to contain (in addition to benefiting from) the hsp72-dependent burst in viral replication during the acute phase of infection. This relationship would indicate a protective role for fever-induced hsp72 and would further suggest an immune selectional pressure for the emergence of viral variants that are hyporesponsive to hsp72 (e.g., Ed N-522D). The prevalence of the Ed N-522D sequence variant supports this possibility, particularly since the N522D substitution diminishes viral fitness (9). Conversely, elevated hsp72 levels can be detrimental to the host if the host is infected with an hsp72-responsive virus, and adaptive immune responses cannot contain the hsp72-dependent increase in viral gene expression during the acute phase of infection. Either scenario defines hsp72 and the C terminus of the MV N protein as significant determinants of viral neurovirulence, with the host beneficial versus detrimental role being context dependent. Ongoing studies will directly test this model by generating NSE-hsp72^+/+ mice on an H-2^d (resistant) background.

It is likely that the effects of hsp72 on MV pathogenicity outside of the CNS would be as disparate as its effects within the CNS, with the outcome dependent on the host immune status. MV exhibits hsp72/heat shock responsiveness that is similar between continuous cell lines of epithelial/mesenchymal origin (Vero, HEp-2, and 293 cell lines) (40, 48, 53, 54) and those of CNS origin (human astrocytoma and murine neuroblastoma) (9, 49). Additional variables that must be considered are potential hsp72-dependent effects of MV on antiviral immunity and the effect that increases in viral gene expression and genome replication may have on cell-free infectious viral progeny release. Ultimately, it is the impact of hsp72 on peripheral cell-free infectious viral progeny release that will define the degree to which MV has exploited a host defense mechanism (i.e., hsp72 induction) in support of viral transmission (i.e., enhanced peripheral virus spread and shedding from mucosal surfaces).