Aberrant infection and persistence of varicella-zoster virus in human dorsal root ganglia in vivo in the absence of glycoprotein I

Abstract

Varicella-zoster virus (VZV) causes varicella, establishes latency in sensory ganglia, and reactivates as herpes zoster. Human dorsal root ganglia (DRGs) xenografts in immunodeficient mice provide a model for evaluating VZV neuropathogenesis. Our investigation of the role of glycoprotein I (gI), which is dispensable in vitro, examines the functions of a VZV gene product during infection of human neural cells in vivo. Whereas intact recombinant Oka (rOka) initiated a short replicative phase followed by persistence in DRGs, the gI deletion mutant, rOkaΔgI, showed prolonged replication with no transition to persistence up to 70 days after infection. Only a few varicella-zoster nucleocapsids and cytoplasmic virions were observed in neurons, and the major VZV glycoprotein, gE, was retained in the rough endoplasmic reticulum in the absence of gI. VZV neurotropism was not disrupted when DRG xenografts were infected with rOka mutants lacking gI promoter elements that bind cellular transactivators, specificity factor 1 (Sp1) and upstream stimulatory factor (USF). Because gI is essential and Sp1 and USF contribute to VZV pathogenesis in skin and T cells in vivo, these DRG experiments indicate that the genetic requirements for VZV infection are less stringent in neural cells in vivo. The observations demonstrate that gI is important for VZV neurotropism and suggest that a strategy to reduce neurovirulence by deleting gI could prolong active infection in human DRGs.

Varicella-zoster virus (VZV) is a human alphaherpesvirus with a DNA genome with at least 74 ORFs that encode viral proteins. VZV causes varicella and establishes latency in sensory nerve ganglia (1). After inoculation of respiratory epithelial cells, VZV reaches skin through a T cell-associated viremia (1, 2). VZV may gain access to sensory neurons by retrograde axonal transport from skin lesions or by viremia. VZV reactivation from latency causes herpes zoster by anterograde axonal transport of virions to skin (1, 3).

VZV pathogenesis is difficult to investigate because of its restricted host range. Establishing xenografts in mice with SCID immunodeficiency permits analysis of VZV tropism for human skin, T cells, and dorsal root ganglia (DRGs) in vivo (4, 5). VZV replicates in DRG xenografts but undergoes a unique transition to persistence within 28 days, in contrast to progressive lytic infection in skin and T cell xenografts. VZV persistence in DRGs is marked by cessation of virion formation, low VZV genome copies, and transcription of immediate-early (IE) genes, especially ORF63, but not ORF31, which encodes the late glyco protein, gB (5). These observations are consistent with studies of human autopsy DRGs in which IE and early gene transcripts are detected, with ORF63 transcripts being most abundant (6–9).

In these experiments, the role of glycoprotein I (gI) in VZV neuropathogenesis was examined by inoculating DRG xenografts with a gI deletion mutant or mutants in which gI expression was reduced by substitutions in the gI promoter to disrupt binding of the cellular transactivator, specificity factor 1 (Sp1), or both Sp1 and upstream stimulatory factor (USF) (10–12). gI is a type I membrane protein encoded by ORF67 that forms heterodimers with the major VZV glycoprotein, gE (10, 13–16). VZV gI deletion mutants replicate in human diploid cells in vitro, although gE trafficking to plasma membranes, cell–cell spread, and syncytium formation are disrupted (10, 12, 17). Deleting gI had no consequences in a rodent model of VZV neuropathogenesis, but this host is nonpermissive (18). The gI homologues in herpes simplex virus 1 and pseudorabiesvirus, along with gE, influence anterograde transport, neuron–neuron spread, and neurovirulence in animal models (19–24). Although dispensable in vitro, VZV gI is required for replication in SCIDhuman (SCIDhu) skin and T cell xenografts (25), and the Sp1 and USF gI promoter mutants have impaired VZV skin and T cell tropism (12). Investigating these VZV mutants in DRGs permitted an assessment of genetic requirements for VZV neurotropism and a comparison of gI and cellular factors that modulate gI expression as determinants of virulence in human neural cells, epidermal cells, and T cells in vivo (12, 25).

VZV is the only herpesvirus for which vaccines, made from the Oka strain, are licensed (26, 27). Oka vaccines are attenuated in healthy children and adults, but not in immunocompromised patients (28), and vaccine Oka is as infectious as parent Oka in SCIDhu DRG xenografts (5). Therefore, defining the contributions of VZV genes and promoter elements to neuropathogenesis has potential relevance for the design of further attenuated VZV vaccines.

Results

Replication and Histopathological Changes.

When DRGs were inoculated with the intact recombinant VZV, rOka, or with rOkaΔgI, from which ORF67 was deleted (10), infectious virus was recovered at day 14 after infection from 7 of 11 rOka-infected DRGs compared with 1 of 11 rOkaΔgI-infected DRGs. rOka was not isolated from any of five DRGs at day 28 or six DRGs at day 56, whereas two of four DRGs inoculated with rOkaΔgI yielded infectious virus at day 70 [supporting information (SI) Table 1]. Compared with uninfected DRGs (Fig. 1A), DRGs exhibited disorganized architecture at day 14 after rOka inoculation; many neurons had a contracted cytoplasm (Fig. 1B, black arrow) and some satellite cells were small and irregular (Fig. 1B, white arrow). DRG architecture did not differ from uninfected DRGs at day 14 after infection with rOkaΔgI (Fig. 1C), but some neurons showed cytopathic changes at day 70 (Fig. 1D, white arrows), which is consistent with prolonged production of virus.

Fig. 1.

Histopathological effects of rOka, rOkaΔgI, and gI promoter mutants in DRGs. Ten-micrometer sections of DRGs were stained with H&E or anti-VZV antibodies. (A) Uninfected DRGs, 17 weeks after transplantation, H&E stain. (B) rOka-infected DRGs at day 14 after infection stained with H&E. The black arrow indicates abnormal neuronal morphology, and the white arrow indicates abnormal satellite cell morphology. (C) rOkaΔgI-infected DRGs at day 14 after infection stained with H&E; no cytopathic changes. (D) rOkaΔgI-infected DRGs at day 70 after infection stained with H&E. The white arrows indicate cytopathic neurons. (E) rOka-gI-SP1/USF-infected DRGs at day 14 after infection stained with rabbit anti-IE63-labeled antibody (detected by using a Texas red-conjugated secondary antibody), anti-human polyclonal antibody (detected by using an FITC-conjugated secondary antibody), and Hoechst (blue) nuclear counterstain. (F) rOka-gI-SP1/USF-infected DRGs at day 14 after infection stained with H&E. The white arrow indicates the contracted cytoplasm of an infected neuron.

VZV Genome Synthesis and Viral Gene Transcription.

rOka-inoculated DRGs (6.1 × 10⁵ pfu/ml) had a mean of 4.2 × 10⁷ copies per 10⁵ cells at day 14, which declined to 2.5 × 10⁶ copies per 10⁵ cells at day 28 by ORF31 PCR (P ≤ 0.05), as expected for the transition to VZV persistence in DRGs. DRGs inoculated with rOkaΔgI (9.0 × 10⁴ pfu/ml) had 1.1 × 10⁵ copies per 10⁵ cells at day 14 and 1.4 × 10⁵ copies per 10⁵ cell at day 28 (P > 0.05). When the expected decrease in genome copy number was not observed with rOkaΔgI, the experiment was repeated with a longer replication interval. In this experiment, DRGs infected with rOka (1.8 × 10³ pfu/ml) had a mean of 5.7 × 10⁷ copies per 10⁵ cells at day 14, declining to 1.4 × 10⁶ copies per 10⁵ cells at day 56 (P ≤ 0.05) by ORF31 PCR. DRGs inoculated with rOkaΔgI (3.7 × 10³ pfu/ml) had 1.9 × 10⁵ copies per 10⁵ cells at day 14, which increased to 2.6 × 10⁶ copies per 10⁵ cells at day 56 (P ≤ 0.05) (Fig. 2A). The same patterns were observed in rOka- and rOkaΔgI-infected DRGs tested by ORF62 PCR (SI Fig. 6A).

Fig. 2.

VZV genome copies and gene transcription in DRGs infected with rOka, rOkaΔgI, or gI promoter mutants. (A) VZV genome copy numbers were assessed by quantitative real-time PCR by using probes ORF31 (A) and ORF62 (see SI Fig. 6A) at day 14 (black bars) and day 56 (gray bars) after infection of DRGs with rOka or rOkaΔgI; three to five DRGs were tested at each time point. VZV genome copy numbers are reported as mean copies per 10⁵ human cells ± standard error. (B) VZV genome copy numbers were assessed by quantitative real-time PCR by using probes for ORF31 (B) and ORF62 (see SI Fig. 6B) at day 14 (black bars) and day 28 (gray bars) after infection of DRGs with rOka or the gI promoter mutants, rOka-gI-Sp1 and rOka-gI-Sp1/USF; three to five xenografts were tested at each time point. VZV genome copies are reported as mean copies per 10⁵ human cells ± standard error. (C and D) VZV transcripts for ORF31 (gB), ORF62, and ORF63 were assessed by quantitative real-time RT-PCR at day 56 after infection in 12 VZV-infected DRGs (C) and in rOka (D Left) and rOkaΔgI (D Right) infected DRGs in a separate experiment. Dots represent individual DRGs. Results are reported as mRNA copies per nanogram of human mRNA (hypoxanthine phosphoribosyltransferase control). DRG samples that contained <0.1 ng of huRNA per well or <50 copies per DRG were outside the limits of the standard curve and were excluded.

When ORF31, ORF62, and ORF63 transcripts were measured in VZV-infected DRGs (Fig. 2C), ORF31 (gB) transcripts were not detected in any of 12 DRGs at day 56, whereas ORF62 transcripts (440 ± 126) and more abundant ORF63 transcripts (2,646 ± 677) were detected in all DRGs, confirming that VZV persistence is characterized by a limited transcriptionally active state. rOka-infected DRGs showed no detectable ORF31 mRNA but low levels of ORF62 and ORF63 mRNA at day 56 (Fig. 2D Left). In contrast, rOkaΔgI-infected DRGs had high levels of ORF31, ORF62, and ORF63 transcripts (ORF31, 32,640 ± 10,246; ORF62, 26,445 ± 6,702; ORF63, 102,359 ± 25,767) (Fig. 2D Right). All rOka and rOkaΔgI comparisons were significant (P ≤ 0.05).

VZV DNA and Viral Proteins.

VZV DNA in situ hybridization (ISH) and expression of IE62, IE63, gE, or synaptophysin, a membrane glycoprotein in neuron vesicles, were examined in thin sections of DRGs infected with rOka or rOkaΔgI (Fig. 3). VZV DNA and IE63 expression were observed in rOka-infected DRGs at day 14 in neuronal nuclei (Fig. 3B, large arrow) and satellite cells (Fig. 3B, small arrow). The VZV DNA signal was also detected along the border of the neuronal cell cytoplasm where viral particles accumulate (Fig. 3B) but was not detected in the negative control (Fig. 3A). At day 56, VZV DNA was detected at the margins of neuronal nuclei (Fig. 3 C, control, vs. D). Because of its delayed replication, rOkaΔgI-infected DRGs were examined at day 42. VZV DNA was detected in nuclei of neurons (Fig. 3E, long arrow) and satellite cells (Fig. 3E, short arrow) but the ISH signal was less intense than in rOka-infected DRGs. Many cells that expressed IE63 had no VZV DNA signal in rOkaΔgI-infected DRGs (Fig. 3F), suggesting fewer genome copies, although better detection of IE63 with the high potency anti-IE63 antibody cannot be excluded.

Fig. 3.

VZV DNA and protein expression in DRGs infected with rOka or rOkaΔgI by confocal microscopy. (A–D) Thin sections (100 nm) of rOka-infected DRGs removed at day 14 (A and B) and day 56 (C and D) were evaluated for VZV DNA localization by ISH using VZV DNA probe (B and D) or a control probe (A and C). The DIG-labeled probe was detected by using an anti-DIG monoclonal with FITC-conjugated secondary antibody (green), and sections were evaluated by confocal microscopy. ISH combined IE63 staining using anti-IE63 rabbit polyclonal antibody (B) or antibody against the neuron-specific protein synaptophysin by using an anti-synaptophysin rabbit polyclonal antibody (C and D); the secondary antibody was Texas red-conjugated anti-rabbit IgG (red). (E and F) DRGs infected with rOkaΔgI were evaluated at day 42 after inoculation by using combined ISH with the anti-DIG VZV DNA probe with anti-DIG FITC-conjugated secondary antibody and staining with anti-IE63 antibody/Texas red secondary antibody. (B and E) The long white arrow indicates hybridization with the VZV DNA probe in a neuron, and the short white arrow indicates the hybridization signal in a satellite cell in the same DRG section. (F) Some IE63-expressing cells did not contain detectable VZV DNA (long white arrows). (G and H) rOkaΔgI-infected DRG sections at day 42 stained with anti-IE63 antibody and Texas red-conjugated secondary antibody without ISH. (I and J) rOka-infected DRG sections at day 14 (I) or rOkaΔgI at day 42 (J) stained with anti-IE63 with Texas red-conjugated secondary antibody and anti-gE monoclonal antibody with FITC-conjugated secondary antibody (green). (I) The short white arrows outline the cell membrane. Images show merged channels.

In rOkaΔgI-infected DRGs, regions of the same section had IE63 expression in most cells (Fig. 3H), whereas in other regions IE63 was detected in only one to two cells (Fig. 3G). The presence of single VZV-infected neurons in rOkaΔgI-infected DRGs, surrounded by uninfected neurons and satellite cells, suggests that gI is required for efficient cell–cell spread. Because gI influences gE trafficking, gE expression was also examined. In rOka-infected DRGs, gE was localized along plasma membranes with minimal cytoplasmic staining in cells expressing IE63 (Fig. 3I). In contrast, in rOkaΔgI-infected DRGs, gE was restricted to the cytoplasm in a punctate pattern that resembles the distribution of Golgi stacks (Fig. 3J, arrow); IE63 in a neuron surrounded by IE63-negative satellite cells again suggests limited cell–cell spread of rOkaΔgI in DRGs.

VZV protein synthesis was further assessed in rOka and rOkaΔgI-infected DRGs by immunoblot. Bands of the size expected for gE (70–96 kDa) were observed in all of three lysates prepared from DRGs at day 14 after rOka inoculation, but not in three specimens from day 56 (SI Fig. 7A). In contrast, bands of the expected sizes for gE and IE63 were detected in all of three rOkaΔgI-infected DRG xenografts at day 70 (SI Fig. 7 B and C). Uninfected DRGs showed no reactivity to VZV antibodies.

Ultrastructural Analysis of DRGs After rOka and rOkaΔgI Infection.

VZV-infected DRGs were examined by electron microscopy of cryosections (cryoimmuno-EM) using the Tokuyasu method, which permits sensitive immunogold single and double labeling combined with ISH (29, 30). In DRGs examined at day 14 after rOka infection, pleomorphic nucleocapsid structures (Fig. 4A, white arrow) were found in nuclei in which IE62 protein was also detected [Fig. 4A, black arrow, 5-nm gold particle (gp)]. VZV DNA was detected within cell nuclei by ISH (Fig. 4C, arrow, 15-nm gp). Unlike IE62, which was distributed throughout nuclei, the VZV DNA signal was localized to discrete nuclear compartments in association with electron dense regions (Fig. 4C). This is evidence that VZV DNA molecules are synthesized in specialized replication compartments within the nuclei of differentiated human cells in vivo. Whereas varicella-zoster (VZ) virions were found almost exclusively in neurons and only rarely in satellite cells by standard transmission electron microscopy (TEM) (5), VZV DNA and VZ virions were detected in both neurons and satellite cells by cryoimmuno-EM, which better preserves membranes between neurons and adjacent satellite cells (SI Fig. 8). Intracytoplasmic vesicles containing virions were observed in satellite cells and neurons (SI Fig. 8), indicating that both cell types can support VZV replication.

Fig. 4.

Comparative ultrastructural analysis of rOka- and rOkaΔgI-infected DRGs by cryoimmuno-EM. DRGs were evaluated at day 14 (rOka-infected) (Left) and day 42 (rOkaΔgI-infected) (Right) by cryoimmuno-EM coupled with VZV protein labeling using gold-conjugated secondary antibodies. (A and B) Viral nucleocapsids (white arrow) and labeling with rabbit anti-IE62 and 5-nm gold-conjugated anti-rabbit Ig (black arrow). (C and D) VZV DNA ISH (black arrows, 15-nm gp). (E and F) Enveloped VZ virions in the cytoplasm labeled with IE62 (10-nm gp). See SI Fig. 8 for larger images.

Overall, few virions were observed in DRGs infected with rOkaΔgI compared with rOka-infected DRGs. VZV DNA was localized to intranuclear zones within electron-dense regions as in rOka-infected cells (Fig. 4D, black arrows, 15-nm gp), but at day 42, only 37 of 315 nuclei (12%) contained VZV DNA in rOkaΔgI-infected DRGs compared with 52 of 104 nuclei (50%) in rOka-infected DRGs at day 14. However, within infected cells, nucleocapsid structures (Fig. 4B, white arrow), nuclear IE62 (Fig. 4B, black arrow) and cytosolic virions (Fig. 4F) were indistinguishable from rOka-infected cells. Both rOka and rOkaΔgI virions had an electron-light core, an electron-dense tegument that contained IE62 (Fig. 4 E and F, white arrow, 5-nm gp), and a bilayered envelope with a halo of envelope proteins. In addition, VZV gE was detected on rOkaΔgI virion envelopes, despite the absence of gI (Fig. 5A, 10-nm gp). Thus, rOkaΔgI replication in DRGs, although restricted and delayed, was not associated with any apparent defects in VZ virion ultrastructure or incorporation of IE62 and gE.

Fig. 5.

Cryoimmuno-EM analysis of rOkaΔgI virus. (A) Normal VZ particles within a vesicle in rOkaΔgI-infected DRGs, with gE on the envelope (15-nm gp). (B) Unusual-appearing membrane stacks within electron-dense region, showing gE on the membrane (white arrows indicate white lines). (C) Altered Golgi structures, with gE immunoreactivity. (D) gE retention (black arrows, gp) on rough endoplasmic reticulum membranes (white arrows) in rOkaΔgI-infected DRGs.

Absence of gI Restricts gE Trafficking and Alters Golgi Cisternae.

Although virion ultrastructure was similar, rOka- and rOkaΔgI-infected neurons showed significant differences in intracellular localization of VZV particles and subcellular organelles (Fig. 5). Cytoplasmic enveloped virions containing IE62 were observed in rOka-infected cells at day 14 (Fig. 4E and SI Fig. 8G) and accumulated within large secretory vesicles, presumed to be late endosomes–prelysosomes (SI Fig. 8G). These vesicles contained pleomorphic particles, some of which appeared degraded. Similar structures were observed in rOkaΔgI-infected DRGs, and contained viral particles that incorporated gE. However, the cytoplasm of rOkaΔgI-infected cells also had distorted Golgi-like structures that were associated with electron-dense regions and incorporated gE (Fig. 5 B and C, 15-nm gp). These Golgi-like structures contained alternating light (membranes) and dark regions (presumably tegument proteins) within Golgi stacks; the gE signal was directly associated with the Golgi membranes (light regions). gE was also detected within the rough endoplasmic reticulum in rOkaΔgI-infected cells, but not in rOka-infected cells, indicating that gE is retained in the rough endoplasmic reticulum of neurons in the absence of gI (Fig. 5D, black arrows, 15-nm gp)

gI Promoter Mutations.

DRGs were infected with rOka-gI-Sp1, in which the binding site for Sp1 is disrupted or with rOka-gI-Sp1/USF, which has mutations of both Sp1 and USF motifs (11, 12). The virulence of rOka-gI-Sp1 was impaired in skin and T cell xenografts, whereas rOka-gI-Sp1/USF did not replicate in skin and was severely limited in T cells (12). Fibroblasts infected with rOka-gI-Sp1 and rOka-gI-Sp1/USF had 5.4- or 7.1-fold, respectively, less gI than rOka-infected cells by immunoblot (SI Fig. 7E). When DRGs were inoculated with rOka (6.1 × 10⁵ pfu/ml), rOka-gI-Sp1 (4.3 × 10⁵ pfu/ml), or rOka-gI-Sp1/USF (1.1 × 10⁶ pfu/ml), infectious virus was recovered at day 14 from three of three DRGs infected with rOka, two of three infected with rOka-gI-Sp1, and three of three infected with rOka-gI-Sp1/USF (SI Table 1). Cytopathic changes induced by the gI promoter mutants were equivalent to rOka-infected DRGs at day 14 (Fig. 1 F vs. B). IE63 and gE expression was similar to rOka-infected cells by confocal microscopy (Fig. 1E). gI expression in rOka-gI-Sp1- and rOka-gI-Sp1/USF-infected DRGs was equivalent to rOka-infected DRGs by immunoblot (1.4 ± 0.04 OD, rOka; 1.4 ± 0.2 OD, rOka-gI-Sp1; 1.4 ± 0.08 OD, rOka-gI-Sp1/USF) (SI Fig. 7 F and G). IE4 protein expression was also equivalent (SI Fig. 7D).

At day 14, mean VZV genome copies detected by using the ORF31 probe were similar in DRGs infected with rOka (4.2 × 10⁷ copies per 10⁵ cells), rOka-gI-Sp1 (6.7 × 10⁷ copies per 10⁵ cells), or rOka-gI-Sp1/USF (2.5 × 10⁸ copies per 10⁵ cells) (Fig. 2B). The decrease observed in rOka-infected DRGs at day 28 also occurred in DRGs infected with rOka-gI-Sp1 and rOka-gI-SP1/USF (day 14 vs. day 28; P ≤ 0.05). Results were equivalent by using the ORF62 probe (SI Fig. 6B). VZV genome copies in rOka-infected DRGs and those infected with the gI promoter mutants were not significantly different when compared at day 14 or 28 by using either probes (Fig. 2B and SI Fig. 6B).

Discussion

These investigations demonstrated that the VZV membrane glycoprotein, gI, is important for infection of human DRGs in vivo. Without gI, VZV genome copies were 100-fold lower at 14 days after inoculation, only 9% of DRGs yielded infectious virus compared with 67% of those infected with intact VZV, and only ≈5–10% of cells within DRGs were infected. Productive rOka infection stopped within 28 days, as was observed in DRGs inoculated with VZV parent Oka or vaccine Oka (5), whereas the absence of gI caused a prolonged infectious process in DRG xenografts during which VZV DNA synthesis, transcription and translation of IE and late genes, and production of infectious virions continued for at least 70 days without the early transition to persistence that occurs in DRGs infected with intact VZV.

The ultrastructural analysis of virions in DRG neurons and satellite cells showed that the absence of gI was associated with the formation of few virus particles and unusual Golgi stacks in the cytoplasm. Without gI, VZV infection of neural cells appeared to be impaired at a late stage of the viral life cycle, reducing secondary envelopment, virion egress, and cell–cell spread. However, if DRG cells become infected, viral DNA replication and viral gene transcription and translation and nucleocapsid assembly occurred. In contrast, deleting gI is completely lethal for VZV infection of skin and T cell xenografts, which show no detectable VZV DNA or virion synthesis after rOkaΔgI inoculation (12, 25). Although quite inefficient, some infectious virions were produced in DRGs in the absence of gI. Thus, these experiments provide evidence that the viral genetic requirements for VZV replication in the DRG microenvironment are significantly less stringent than in differentiated human epidermal and T cells in vivo.

Altered intracellular trafficking of gE is likely to be the mechanism for impaired VZV infection of DRGs when gI is lacking. VZV gE and gI form heterodimers in infected cells and gI contains an endocytosis motif that potentiates gE retrieval from plasma membranes and trafficking to sites of virion envelopment (16, 31). In cultured cells infected with gI-null mutants, gE cell surface expression is diminished, posttranslational modification of gE is abnormal, irregular membranous structures and vacuoles accumulate in the cytoplasm, and virion egress to cell surfaces is altered (10, 14–16, 32). Similarly, neural cells infected with rOkaΔgI had little gE on cell surface membranes and distorted Golgi structures. Rough endoplasmic reticulum retention of gE and gE association with distorted Golgi structures, demonstrated by immunogold electron microscopy, implicate aberrant gE intracellular trafficking as a mechanism for impaired replication of rOkaΔgI in DRGs. VZV gE enhances cell–cell contact in polarized epithelial cells (33) and colocalization of VZV gE–gI with host cell membrane tight junction proteins suggests that gE–gI may facilitate particle movement across intracellular junctions, which is a function of gE orthologues in other neurotropic alphaherpesviruses (21, 34, 35). Abnormal gE trafficking in the absence of gI could also explain limited cell–cell spread in DRGs because gE–gI complexes potentiate polykaryon formation (36). In addition, gE interaction with the insulin-degrading protein, which has been identified as a gE receptor in infected cells in vitro and which is present in human neurons, could be disrupted (37). Nevertheless, because VZV gE is essential, some gE must reach sites for virion envelopment independently of gI in order for VZV replication to occur in differentiated human neural cells, and, despite aberrant gE trafficking, some VZ virions that expressed gE were produced in DRGs in the absence of gI.

The promoter elements of herpesvirus genes have critical binding sites for cellular transactivators that modulate transcription in combination with viral transactivating proteins. Optimal gI promoter function in vitro requires an activating upstream sequence containing cellular transcription factors binding motifs (11). As demonstrated by mutagenesis of binding sites for Sp1 and USF in the promoters of VZV ORF67 (gI), ORF10 and ORF68 (gE), these cellular factors can influence VZV virulence in skin and T cell xenografts in vivo (12, 38, 39). The replication of rOka-gI-Sp1 was impaired in skin and T cells and rOka-gI-Sp1/USF did not infect skin (12). In contrast, disrupting the Sp1 and USF binding sites in the gI promoter had no consequences for gI expression in neuronal tissue, nor VZV neurotropism, either during acute infection or persistence. Other cellular transactivating factors may interact with IE62 to regulate gI expression, or IE62 alone is sufficient, in DRG cells. These observations document tissue-specific differences in the requirements for cellular transactivators known to enhance herpesviral gene expression from promoter construct experiments in vitro (11). VZV neurotropism appears to be less dependent on Sp1 and USF regulatory activity than infection of differentiated human skin and T cells in vivo.

Current VZV vaccines are attenuated for replication in skin, but not in human T cells and DRG xenografts (4, 5), which is consistent with observations of insufficient attenuation of Oka vaccines in immunocompromised children. The evaluation of recombinants derived from vaccine Oka in SCIDhu DRG xenografts has the potential to identify strategies for constructing a second-generation recombinant vaccine that exhibits neuroattenuation in human ganglia in vivo. The transition from replication to VZV persistence in human neurons in SCIDhu DRG xenografts suggests that, in the absence of an adaptive immune response, an equilibrium is achieved between VZV and innate cellular responses of neurons and/or satellite cells that favors persistence in sensory ganglia. In this context, the diminished but prolonged replication of rOkaΔgI in human DRGs suggests that attempts to attenuate VZV neurovirulence based on targeted mutations that limit viral replication in vitro through glycoprotein or other mutations may have unpredicted consequences for establishing this balance.

Materials and Methods

DRG Xenotransplantation.

Human fetal DRGs were inserted under the kidney capsule of male C.B-17 scid/scid mice (Taconic Farms, Germantown, NY) (5). The Stanford University Administrative Panel on Laboratory Animal Care approved all animal protocols. Human tissues were provided by Advanced Bioscience Resources (Alameda, CA) and were in accordance with state and federal regulations.

Viruses and Inoculation of DRG Xenografts.

Recombinant Oka, rOkaΔgI, rOka-gI-Sp1, and rOka-gI-Sp1/USF viruses were from the same stock as described by Ito et al. (12). Viruses were propagated in human embryonic lung fibroblasts cells; inoculum titers were determined by infectious focus assay at the time of injection (5). DRGs were infected by direct injection of VZV-infected human embryonic lung fibroblasts cells at 4–12 weeks after transplantation (5).

Analysis of Viral Replication, Cytopathic Changes, and VZV DNA/RNA Synthesis.

At designated times after inoculation, mice were killed and DRGs were removed and immersed in 4% paraformaldehyde for histologic or TEM fixation or to prepare finely minced tissue in PBS (5). Minced tissue was used for infectious virus assay and DNA extraction or placed in RNAlater solution (Invitrogen, Carlsbad, CA). To assess virus production, half of the mincate was cocultured with uninfected MeWo cells and passaged until plaques appeared or for at least six times, before being considered negative. The second aliquot of minced tissue was treated with DNAzol or TRIzol (Invitrogen) for DNA or RNA extraction.

DRG Cryosections.

DRGs were fixed in 4% paraformaldehyde in PBS (0.1 M, pH 7.2), postfixed in 0.05% glutaraldehyde and 4% paraformaldehyde for 2 h at 4°C, washed in PBS, and infiltrated with 2.3 M sucrose for 24 h at 4°C. DRGs were mounted on ultramicrotomy pins and frozen in liquid nitrogen; 80- to 100-nm ultrathin cryosections were prepared for TEM, and 150-nm, 500-nm, or 1-μm cryosections were prepared for immunofluorescence. Sections were transferred to formvar/carbon-coated nickel grids and processed on the same day for ISH or immunostaining.

ISH for VZV DNA and Detection of VZV and Cellular Proteins by Immunoblot, Confocal Microscopy, and TEM.

Thin cryosections were denatured in NaOH and washed before incubation with a hybridization mix (Maxim Biotech, San Francisco, CA) containing the digoxigenin (DIG)-labeled VZV DNA (pBR322 vector containing the HindIII C fragment of VZV) or a negative control probe (pBR322 vector alone). The probes were labeled by random priming using a DIG DNA labeling kit (Roche Diagnostics, Penz berg, Germany). After overnight hybridization at 37°C, sections were washed, blocked, and processed for ISH; nuclear DNA was counterstained with Hoechst. For electron microscopic analysis, sections were incubated with anti-DIG monoclonal antibody (Roche) for 1 h in DIG-blocking buffer (Roche), washed, and incubated with an anti-mouse polyclonal antibody conjugated with 5-, 10-, or 15-nm colloidal gold. Sections were analyzed in a JOEL TEM. Sections that were examined by immuno-double-labeling were incubated with antibody to the protein of interest, which was detected with a secondary anti-rabbit Alexa 647-conjugated antibody (Molecular Probes, Eugene, OR) for immunofluorescence or with a rabbit anticolloidal gold-conjugated antibody (Ted Pella, Redding, CA) for TEM. For DNA in situ signal quantification, three grids per virus were counted, and each grid contained 30 sections. A gold cluster was considered to represent a bona fide signal when it was exclusively nuclear and not detected in any extracellular structures.

Individual DRG cell lysates for immunoblot were prepared in radioimmunoprecipitation assay buffer (Boston BioProducts, Worcester, MA), were separated by SDS/PAGE, and were transferred to PVDF membranes (Millipore, Bedford, MA). VZV proteins were detected by using ECL reagents (Amersham, Buckinghamshire, U.K.). Membrane-exposed film was imaged by using a GS-710 instrument (Bio-Rad, Hercules, CA) and analyzed by using Quantity One software (Bio-Rad).

Antibodies.

Antibodies included rabbit polyclonal anti-IE4 and anti-IE62 (gifts from P. Kinchington, University of Pittsburgh, Pittsburgh, PA), a rabbit polyclonal anti-IE63 (a gift from W. Ruyechan, University of Buffalo, Buffalo, NY), and a rabbit anti-gI antibody (a gift from S. Silverstein, Columbia University, New York, NY). The anti-IE63 and anti-gI antibodies were protein G column-purified IgG (Pierce, Rockford, IL). The mouse monoclonal anti-gE antibody (Chemicon, Temecula, CA) and rabbit anti-human synaptophysin antibody (Zymed, South San Francisco, CA) are commercially available.

Quantitative PCR and RT-PCR.

Quantification of VZV genome copies and quantitative RT-PCR for viral transcripts was done by using the 5′-exonuclease method as described previously (5). VZV genome copies were determined per 10⁵ cells by using probes for both ORF31 (gB) and ORF62, which is present as two copies per VZV genome, in each assay. Samples with <0.1 ng of human RNA per well were excluded because they were outside the linearity limits of the hypoxanthine phosphoribosyltransferase standard curve.

Statistical Analysis.

Comparisons were made by using the Student t test (Excel software, version 11.2; Microsoft, Redmond, WA).

Abbreviations

VZV

varicella-zoster virus

VZ

varicella-zoster

DRG

dorsal root ganglion

gI

glycoprotein I

Sp1

specificity factor 1

USF

upstream stimulatory factor

r

recombinant

ISH

in situ hybridization

IE

immediate early

gp

gold particle

cryoimmuno-EM

electron microscopy of cryosections

TEM

transmission electron microscopy

DIG

digoxigenin.