Cellular heat shock proteins (HSPs) are one of the most phylogenetically conserved classes of proteins with critical roles in maintaining cellular homeostasis and in protecting the cells from stressful conditions, reflecting in large part their ability to facilitate folding of nascent protein or refolding of denatured protein [1]. HSPs are present in a wide variety of organisms, tissues, and cells, and occupy diverse intracellular locations including the cytosol of prokaryotes, and the cytosol, nucleus, endoplasmic reticulum, mitochondria and chloroplasts of eukaryotes [2]. HSPs are classified into families according to their mass. These include small HSP, HSP40, HSP60, HSP70, HSP90, and HSP110. Each HSP family contains proteins with closely related structures, whereas HSP families show little obvious amino acid homology to one another [3].

Since their initial discovery, diverse functions of HSPs have been elucidated. As molecular chaperones, HSPs maintain protein conformation, facilitate trafficking and assembly/disassembly of multimeric complexes [1]. But HSP may also bind native protein, altering their conformation and thus function. This type of interaction is known as HSP-mediated activity control and is in contrast to the stability control associated with HSP-mediated protein folding events [4].

While HSPs are host protective in times of cell stress, they also support gene expression of DNA and RNA viruses, particularly for viruses interacting with the major inducible 70kDa HSP isoform (hsp70). Hsp70 is present in all organisms, ranging from archaebacteria and plants to humans [2,5]. Expression levels of hsp70 vary, with hsp70 being induced and redistributed within cells in support of virus replication and being induced in tissues by febrile temperatures that frequently accompany viral infection. As such, levels of hsp70 may determine infection phenotype through direct support of viral gene expression. Hsp70 also plays an immune modulatory role, highlighting potential to influence innate and adaptive immune response against viral infection. The focus of this review will therefore examine the potential of virus-hsp70 interaction to influence outcome of infection, taking into consideration the role of hsp70 in virus replication as well as innate immune responses. Particular emphasis will be placed on infections of the central nervous system, highlighting animal models that have provided the first insights into the biological significance of virus-hsp70 interaction.

Members of the 70 kDa HSPs family (HSP70) are highly conserved across species. The human HSP70 family includes at least eight unique proteins that differ from each other by level of expression and cellular localization. Hsp70-5(Grp78) and hsp70-9(Grp75) are located in the lumen of the endoplasmic reticulum and in the mitochondrial matrix, respectively. The remaining six HSP70 proteins are hsp70-1a, hsp70-1b, hsp70-1t, hsp70-2, hsp70-6, and hsc70, all of which reside in the cytosol and nucleus. Cytosolic HSP70 family members include a constitutively expressed isoform, hsc70 (or hsc73) and two major stress-inducible members, hsp70-1a and 1b that differ by only two amino acids and are collectively known as hsp70 (also referred to hsp72 in order to facilitate distinction from the family as a whole). For both humans and mice, hsp70-1a and hsp70-1b are encoded by two closely linked, intronless stress-inducible genes (HSPA1A and HSPA1B in humans) located in the major histocompatibility (MHC) class III cluster [6]. There is 50% amino acid identity between human hsp70 and the prokaryotic homologue DnaK, and 76% identity between human hsp70 and yeast SSA. The conservation of hsp70 sequence reflects similar functional properties across species. Rodent hsp70 can be functionally complimented by human hsp70 to induce cellular protection against various stress responses [7,8].

Overexpression of hsp70 increases resistance and protection against heat, ischemia, and oxidative stress [9,10,11,12,13]. Protection is attributed to the chaperone functions of hsp70 that support protein folding, translocation, and assembly reactions [14,15,16,17] and prevention of denatured protein aggregation [1,18]. These same functions would support maturation of high levels of viral protein typical of lytic infections. Hsp70 activity control functions include inhibition of apoptosis. Hsp70 inhibits the signal transduction pathways of apoptosis, including Bax translocation into mitochondria, release of cytochrome c from mitochondria, apoptosome assembly through interactions with Apaf-1, and activation of initiator caspase 3 [19]. Accordingly, inhibition of hsp70 expression in human breast tumor cells result in massive cell death [20]. Inhibition of apoptosis may enhance viral replication by preserving the viral factory. Conversely, adenovirus infection causes a dramatic suppression of hsp70 mRNA levels during the late phases of the viral replication cycle when the viral particles are already assembled, enhancing apoptosis in order to promote viral particle release [21].

Hsp70 can be released from cells following cell death [62], but also via active processes independent of cell death [63,64]. Release following lytic infection has been shown with parvovirus infection of a tumor cell line, where cell death released hsp70 in addition to other heat shock protein family members including hsp60 and hsp90 [65]. We have shown similar virus-induced release of hsp70 in mouse neuroblastoma cells lytically infected with MeV and vesicular stomatitis virus (VSV), where the concentration of intracellular hsp70 is proportionate to the level of hsp70 release (unpublished observation). VSV infection of neuroblastoma cells is highly cytopathic (i.e., lytic). MeV infection of neuroblastoma cells is typically non-cytopathic, although we were able to create a lytic infection phenotype by transfecting cells to express a receptor used preferentially by the infecting strain of MeV, in this case the CD46 receptor and the Edmonston strain of MeV. Lytic infections of control neuroblastoma cells resulted viral induction of hsp70 and low level release. Release was dramatically increased in cells that were stably transfected to constitutively express hsp70.

We have also demonstrated hsp70 secretory release from non-cytopathic MeV infection of neuroblastoma cells (unpublished observation). Secretory release of hsp70 is well documented in exosomal pathways. Exosomes are 40 to 100 nm endosome-derived vesicles released from various cell types under both normal and pathopysiological conditions. They are derived from plasma membranes following endocytosis, where early endosomes are subsequently sorted into multivesicular bodies (MVBs). Inward budding of endosomal MVBs result in generation of intraluminal vesicles (ILVs). There are two distinct classes of MVBs including degradative MVBs which lead to lysosomal degradation and exocytic MVBs, which fuse with the plasma membrane and release ILVs from the cell in the form of exosomes. Exosomes are the only known secreted cellular vesicles derived from internal membranes [66]. Exosomes are shed by various myeloid and lymphoid cells including B and T lymphocytes, dendritic cells, microglia and macrophages, but also tumor cells of multiple lineages and neurons [67]. The molecular composition of exosomes is largely dependent on cell type, with exosomes from various origins containing both conserved proteins and cell type specific proteins. Most exosomes contain cytoplasmic proteins, members of endocytosis pathway proteins, and signal transduction proteins. In addition, heat shock proteins such as hsp70 and hsp90, and MHC class I molecules are commonly identified in exosomes [67]. For example, protein analysis of exosomes secreted from rat cortical neurons showed cytoskeleton, signaling, and membrane trafficking associated proteins including HSP70 family members [68]. Exosomes may contain multiple heat shock proteins under normal conditions, including hsp27, hsp60, hsp70, hsc70, and hsp90, with hsp70 (hsp72) levels being increased by heat stress [69] and intracellular bacterial infection [70]. It was reported that mycobacteria infection of macrophages enhances exosomal release and hsp70 content of exosomes [71]. Hsp70 has been localized to the surface of exosomes, placing it in a position to trigger innate immune signaling by interacting with cell receptors.

Hsp70 release from viable virus-infected cells is a largely unexplored aspect of virus-host interaction. We are currently examining the release mechanism in MeV infected neuroblastoma cells, which may include exosomal release or microvesicular shedding. In vitro infections of human bronchial epithelia cells with RSV also cause hsp70 induction and extracellular release, and tracheal aspirates from RSV infected children showed increased levels of extracellular hsp70, although it was not determined whether the elevation of hsp70 was the result of active secretion or cell death [72]. It was also reported that the level of hsp70 in serum from infected patients was significantly increased relative to healthy control groups [73].

Our group has studied the effects of enhanced hsp70 expression on viral neurovirulence using the mouse model of brain infection. Mice are a good system to study the effects of increased hsp70 levels since the baseline in normal mice is low; mice do not constitutively express hsp70 in neurons (unlike other mammals) and they lack a febrile response following viral infection [58,101]. We can increase hsp70 expression levels by transient hyperthermic treatment or by selective transgenic expression. An additional benefit of the mouse system is that the hsp70 responsiveness of MeV is well characterized in mouse neuroblastoma cells [43] and neurons are a primary target of infection in the mouse model of MeV infection in brain. Intracranial injection of Edmonston MeV to neonatal mice exploits both mouse age and strain-dependent susceptibilities to MeV infection [60,102,103,104,105,106], eliminating the need for transgenic expression of the human membrane cofactor protein (CD46) as an ancillary viral receptor or use of mouse brain adapted strains of MeV in order to promote brain infection [107,108,109,110]. In addition, the high incidence of stable persistent infection and low incidence of mortality are helpful to study the hsp70-mediated effects that may either increase virulence or clearance. Mouse strains show differences in susceptibility to MeV infection that is associated with their H-2 haplotype, major compatibility complex(MHC) class I, II and III loci that determine the efficacy of immune responses to MeV following intracranial challenge [111,112]. The H-2 dependent basis for susceptibility has been related to differential IFN-γ production by MeV specific CD4⁺ T cells. IFN-γ production is enhanced in resistant mice strain carrying the H-2^d allele and diminished in susceptible mice strain carrying the H-2^b allele [113]. The different responses may reflect the activity of immune regulatory genes inducing cytokine expression, since in vitro and in vivo stimulation of lymphocytes from H-2^b mice showed decreased levels of Th1 cytokines resulting in lower amount of IFN-γ production relative to H-2^d mice [114].

Protective immunity in the mouse begins with innate immune responses of uninfected brain macrophages (microglia) to signals from infected neurons [115]. The identity of the signals has not been previously resolved. The key innate responses include production of type 1 interferon, which is predominantly IFN-β in the brain, and cross-presentation of viral antigen on MHC I [116]. MeV infection does not induce IFN-β in neurons [117,118,119], which may reflect a more general restriction of type 1 IFN production in this cell type, although neurons do maintain the capacity to respond to IFN-β by producing antiviral IFN stimulated gene products [117]. MeV also does not induce significant levels of MHC I on neurons, the induction being primarily observed on glia [118,120]. As such, virus-specific T cells primed in the periphery, respond to viral antigen that is cross-presented by activated microglia. These stimulated T cells produce IFN-γ that mediates non-cytolytic viral clearance [115]. The fundamental importance of IFN-β in this process extends beyond the antiviral state induced in neurons to include stimulation of microglial activation [121,122] and stimulation of virus-specific T cells to produce IFN-γ [123,124]. These combined effects make IFN-β expression a key determinant of permissiveness to (or protection against) viral neurovirulence.

Initial studies on the effects of hsp70 on the outcome of MeV brain infection used transient hyperthermic treatment to induce hsp70 in neurons and glia of neonatal Balb/c (H-2^d) mice [58]. Infection followed the hyperthermic treatment. Treated mice showed decreased MeV cytopathic effect, viral burden and increased clearance that was correlated to increased number of virus-specific lymphocytes in spleen relative to control mice. These data suggest that hyperthermic pre-conditioning, with the transient elevation in hsp70 levels, enhance the development of cell-mediated antiviral immunity. Unresolved is the influence of other heat induced stress proteins, or the different roles that heat shock proteins may be playing in virus infected cells relative to the uninfected brain macrophages that are key to innate and adaptive immune responses.

A transgenic model was used to establish the role of virus-hsp70 interaction in the infected neuron to influence innate and adaptive immune responses in uninfected brain macrophages and lymphocytes [61,125]. Transgenic mice that selectively overexpress hsp70 in neurons were created on an H-2^d background. The expression construct was driven by the neuron specific enolase promoter, and the neuron specific expression pattern was confirmed by immunohistochemistry. The hsp70 TG H-2^d mice showed complete protection against Edmonston MeV neurovirulence, compared to 35% mortality in NT controls. Survival analysis supported the statistical significance of the infection outcomes and showed an inverse correlation between survival time and brain viral RNA burden. Included in this study was a genetically engineered Edmonston MeV variant lacking a transcriptional response to hsp70. A single amino acid substitution in the C-terminus of the N protein (N522D) disrupts one of two hsp70 interacting sites, resulting in a selective disruption of hsp70-mediated stimulation of transcription without affecting other aspects of basal or hsp70-dependent replication [43]. Infection of TG mice with the N522D virus resulted in a level of mortality that was not significantly different from that observed in NT mice. Results thus show that a viral transcriptional response to hsp70 is essential to the complete host protection mediated by hsp70.

In order to study the basis for hsp70-mediated host protection, mice infected with MeV were depleted of CD4⁺ and/or CD8⁺ T cells via intraperitoneal injection of CD4 or CD8-specific monoclonal antibodies. Mortality was significantly increased from 35% to 94% in NT mice in which CD4⁺ T cells were depleted, but TG mice in which CD4⁺ T cells were depleted showed only 41% mortality. TG mice lacking both CD4⁺ and CD8⁺ T cells still showed a 27% survival rate. This result shows that hsp70 confers a significant enhancement of innate immunity against MeV-induced mortality. Infected brains were then analyzed at 5 days post infection (d p.i.) by real time RT-PCR of total brain RNA in order to gain insight as to the basis for hsp70-mediated innate immunity. Significant infiltrations of lymphocytes are not observed at 5 d p.i. based upon evaluation of tissue sections and low levels of IFN-γ transcripts in brain. Both lymphocytic infiltrates and IFN-γ was significantly increased at 10 d p.i., indicating the onset of adaptive T cell immune responses. The RT-PCR analysis showed enhanced activation of macrophages in TG relative to NT mice at 5 d p.i. based upon levels of CD68 and MHC II transcripts. We also showed greater levels of TLR2 and TLR4 transcripts in infected TG relative to NT mice.

Results from our laboratory and data in the literature combine support a novel axis of innate antiviral immunity that is driven by hsp70 (Figure 1). In this model, MeV-hsp70 interaction in neurons results in the extracellular release of hsp70. The level of release reflects the intracellular concentration of hsp70 and the magnitude of virus gene expression. The ability of hsp70 to stimulate virus gene expression and for virus gene expression to induce hsp70 thus represents a positive feedback loop driving hsp70 extracellular release. The extracellular hsp70 acts as a damage associated molecular pattern to stimulate innate responses of brain macrophages, particularly the expression of IFN-β. The IFN-β would create an antiviral state in infected neurons, would drive further brain macrophage activation and the cross-presentation of viral antigen derived from infected neurons. Virus specific T cells, encountering viral antigen expressed on activated brain macrophages, would secrete IFN-γ, resulting in viral clearance. IFN-β would be key to this aspect of the adaptive immune response as well. In this model, the extracellular hsp70 acts as a key danger signal, released by infected neurons, to stimulate host protective responses of brain macrophages. The relationship explains how the hsp70-mediated stimulation of virus gene expression is host protective, and not as paradoxical as it may seem. The relationship also provides a mechanistic basis for the host protective effects of fever, given that febrile temperatures are a potent inducer of hsp70 in brain.

Ongoing studies using the MeV model of brain infection are designed to validate the model for hsp70-dependent innate immunity. Unpublished studies detail transcriptome analysis in brain RNA harvested from TG and NT mice at 5 d p.i. The top three canonical pathways that distinguish infected TG from infected NT mice are macrophage maturation, IFN induction and antiviral immunity, and antigen presentation. Results support a greater type I IFN response in TG relative to NT infected mice as the main differences linked to survival. Activation of brain macrophage and antigen presentation is supported by immunochemistry analysis of MHC II, and activation of type I IFN is confirmed by real time RT-PCR. Use of mice lacking a functional type I IFN receptor (IFNAR^−/−) shows the critical role of type I IFN in hsp70-mediated protection. Cell culture studies confirm secretory release of hsp70 from MeV infected neurons and the ability of extracellular hsp70 to serve as a potent stimulus for the production of IFN-β in uninfected microglia.

The mechanism of innate immune activation by hsp70 suggests a broad virological and host relevance. Hsp70 stimulates gene expression of viruses in multiple families, fever is a frequent accompaniment of virus infection, ensuring elevated levels of hsp70 in virus infected cells, and both lytic and non-cytolytic infections can induce extracellular hsp70 release. As a TLR ligand capable of activating macrophages and inducing type 1 IFN, hsp70 can mediate induction of innate immune responses in a far greater sphere of uninfected cells than would be achieved by host responses to a pathogen associated molecular pattern. Work in progress will show this more broad relevance to rhabdoviruses, a family of negative strand RNA viruses that include rabies. Using vesicular stomatitis virus, we have shown that lytic infection of neurons results in high level release of hsp70, and that mice are protected against neurovirulence following a natural route of infection (i.e., intranasal inoculation). Importantly, the protection is observed in weanling mice, proof that the hsp70 protective mechanism is not age-dependent. Although our initial focus has been the central nervous system, the axis of hsp70-mediated innate immunity should also apply to any organ system, reflecting the ubiquitous nature of hsp70 and TLR2/4 expression. Observations with RSV infected children support this contention.

In vitro studies showing that hsp70 stimulate virus gene expression has presented a paradox when considering the phylogenetic conservation of the febrile response to microbial infections. Fever is a potent inducer of hsp70 in all tissue and the conservation of this response suggests a host protective role. The apparent paradox can be reconciled by considering the potential of the increase in virus gene expression to stimulate innate immunity, particularly the expression of type 1 interferons. Historically, focus has been placed upon pathogen associated molecular patterns as the stimulus. The current review highlights the novel role of hsp70 as a damage-associated molecular pattern, released from infected cells and serving as a potent inducer of type 1 interferon. The relative importance of this novel axis of innate immunity is defined by the ubiquitous nature of hsp70 expression, the abundance within virus infected cells, and the apparent dependence of many viruses on hsp70 to support virus replication. The virus-hsp70 relationship is inescapable.