Induction of an Embryonic Mouse Innate Immune Response following Inoculation In Utero with Minute Virus of Mice

ABSTRACT

We used an embryonic-infection model system to show that MVMp, the prototypic minute virus of mice (MVM) serotype and a member of the genus Protoparvovirus, triggers a comprehensive innate immune response in the developing mouse embryo. Direct inoculation of the midtrimester embryo in utero with MVMp results in a widespread, productive infection. During a 96-h infection course, embryonic beta interferon (IFN-β) and IFN-γ transcription were induced 90- and 60-fold, respectively. IFN-β levels correlated with the embryo viral burden, while IFN-γ levels first increased and then decreased. Production of proinflammatory cytokines, interleukin 1β (IL-1β) and tumor necrosis factor alpha (TNF-α), also increased, but by smaller amounts, approximately 7-fold each. We observed increased levels of downstream antiviral effector molecules, PKR and phosphorylated STAT2. Finally, we showed that there is an immune cell response to the virus infection. Infected tissues in the embryo exhibited an increased density of mature leukocytes compared to the same tissues in uninfected embryos. The responses we observed were almost completely restricted to the infected embryos. Uninfected littermates routinely exhibited small increases in innate immune components that rarely reached statistical significance compared to negative controls. Similarly, the placentae of infected embryos did not show any significant increase in transcription of innate immune cytokines. Since the placenta has both embryonic and maternal components, we suggest there is minimal involvement of the dam in the response to infection.

IMPORTANCE Interaction between the small single-stranded vertebrate DNA viruses, the protoparvoviruses, and the host innate immune system has been unclear. The issue is important practically given the potential use of these viruses as oncotherapeutic agents. The data reported here stand in contrast to studies of innate immune response during protoparvovirus infection of adult hosts, which invariably reported no or minimal and sporadic induction of an interferon response during infection. We conclude that under conditions of robust and productive MVM infection, a normal murine host is able to mount a significant and broad innate immune response.

INTRODUCTION

In some respects, the developing mammalian embryo should be paradise for viral pathogens—a rapidly dividing population of potential host cells, expressing all the host metabolic functions the virus needs for its own replication, coupled with a complete absence of adaptive immune responses and an innate response that is considered immature and biased toward tolerance.

Despite this, fetal viral infection is not considered a major cause of complications during pregnancy compared, for instance, to the various bacterial infections causing fetal inflammatory response syndrome (1). It seems unlikely that virus is more efficiently excluded from the embryonic compartment than bacteria (2), and several studies have identified viral signal in a significant fraction of embryos or amniotic fluid from normal pregnancies (3–5). Despite the presence of virus, there is perhaps no increased perinatal risk associated with earlier, asymptomatic infections (4, 6), providing some indication that, despite the lack of an adaptive immune system, the fetus is able to mount an often—but clearly not always— effective antiviral response. In humans, there are a small number of known viral pathogens (e.g., cytomegalovirus [CMV], herpes simplex virus [HSV], HIV, B19 virus, rubella virus, and varicella-zoster virus type 2 [VZV 2]) that can efficiently pass from mother to fetus and mount productive infections leading to persistent infection, fetal abnormalities, or death. There are also sporadic case reports of other viruses doing the same, e.g., influenza virus (7). Searching for viral signatures in cases of idiopathic spontaneous abortions has sometimes (8–10) (and sometimes not [11]) provided evidence that viral infection may be involved.

The Parvovirinae hold a special spot among the viral taxa with respect to mammalian fetal infection. Viruses within almost every subgroup are consistently able to cross the placental barrier and infect the embryo. They include the autonomous parvoviruses (genus Protoparvovirus) (12), e.g., minute virus of mice (MVM) (13) and porcine parvovirus (PPV) (14), the dependovirus adeno-associated virus (AAV) (15), Aleutian mink disease virus (AMDV) (16), and the erythrovirus B19 (17). Viral infection with human erythroparvovirus B19 causes fifth disease in toddlers but is frequently asymptomatic in adults. If B19 infection occurs during pregnancy, it can occasionally lead to severe complications, including fetal anemia and death. We have shown that the rodent parvovirus minute virus of mice leads to a productive widespread infection when inoculated directly into the midgestation mouse embryo (18), although infection in adults is minimal and almost asymptomatic (19).

The extent to which the autonomous parvoviruses interact with the innate immune system is unclear, and the issue continues to chart a complex course in the literature. Earlier studies pointed to minimal or weak induction in animal and tissue culture models (20–23), but two recent independent reports showed induction of innate immune pathways in mouse primary embryonic fibroblasts (MEFs) infected with MVM (24, 25). In a similar experimental system (26), weak and delayed interferon (IFN) production that was insufficient to induce an antiviral state was observed. In contrast, Paglino and colleagues showed that these viruses fail to stimulate a type I interferon response in normal or transformed human cells (27). Independent studies of two protoparvoviruses, Kilham rat virus (KRV) (28, 29) and MVMp (30), provide some evidence for innate immune activation in vitro via Toll-like receptor 9 (TLR9). There are no described mechanisms through which the autonomous parvoviruses may abrogate an innate immune response. However, some preliminary evidence has begun to accumulate (26), and some of the autonomous parvoviruses probably establish persistent infections (31), a hallmark of many viruses that are able to suppress innate immune function.

There are few standard animal models of embryonic viral infection and fetal response. As in humans, there are individual virus-host combinations where the virus routinely crosses the placental barrier to mount a productive, sometimes lethal infection of the embryo, e.g., porcine reproductive and respiratory syndrome virus (32) and porcine parvovirus (33). Of these, it is study of the flavivirus bovine viral diarrhea virus, a relative of human hepatitis C, in its natural host that has produced the most data concerning embryonic immune response to an invading virus (34–38). The data suggest that the mid- to late-gestation bovine fetus is able to mount an interferon antiviral response (37) and that the virus exhibits an ability to suppress the response (39; reviewed in reference 40).

In this work, we describe the development of a new model system for characterizing fetal innate immune response to viral infection and add fresh data concerning interaction between the parvoviruses and the host innate immune system. We show that direct inoculation in utero of the midtrimester mouse embryo with the parvovirus minute virus of mice leads to induction of all the major functional branches of the innate immune pathway in the embryo: antiviral, signaling, and cell mediated.

MATERIALS AND METHODS

Virus preparation.

Infectious wild-type MVMp virus was produced by DOTAP (Roche) transfection of A9 cells, as described previously (41), with an MVMp-bearing infectious plasmid derived from pMM984 (42). Virions were purified from lysates by an iodixanol gradient and quantitated by plaque assay, essentially as described previously (43, 44). Control background material was prepared using the same purification protocol on mock-infected A9 cells and collecting the same iodixanol gradient fractions.

Mouse care.

C57BL/6 mice (Harlan, Rehovot, Israel) were kept in the medical faculty animal care facility. All housing and experimental procedures were reviewed and approved by the Ben Gurion University Committee for the Ethical Care and Use of Animals in Experiments. Females were mated to males and checked the following morning for vaginal plugs. The morning on which a plug was observed was considered 0.5 day postconception (p.c.).

Embryo inoculation in utero.

Virus diluted in sterile physiological saline solution was injected into the livers of 13.5-day p.c. mouse embryos in utero as described previously (18, 45), with the exception that a pulled glass capillary needle and a syringe pump were used instead of a Hamilton syringe for injection.

Embryo dissection and sample preparation.

Dams at chosen times posttreatment, or age-matched unperturbed control dams, were killed, and the embryos with their placentae were removed into cold physiological saline. Only live embryos exhibiting normal morphology were used for subsequent analyses.

Immunohistochemistry.

Embryos were collected, cryosectioned, and subjected to immunohistochemistry essentially as described previously (18, 45). To detect virus, an anti-MVM capsid polyclonal antiserum and a Cy3-labeled secondary antibody (Jackson ImmunoResearch) were used. Control samples collected from uninjected age-matched embryos and from injected embryos at t = 0 and sections from injected embryos processed with secondary antibody alone were used to confirm specificity. To detect mature leukocytes, anti-CD11b rat monoclonal antibody (Abcam; ab8878) and a DyLight550-coupled anti-rat secondary antibody (Abcam; ab98387) were used.

Western blot analysis.

Protein was prepared from embryo homogenates by trichloroacetic acid (TCA) precipitation and then resuspended in cracking buffer (8 M urea, 10 mM NaPO₄, 1% β-mercaptoethanol, 1% SDS). Following addition of sample buffer, samples were heated for 5 min at 95°C and centrifuged, and equal quantities of soluble protein were fractionated by SDS-10% polyacrylamide gel electrophoresis. The proteins were blotted onto a nitrocellulose membrane (Amersham), which was then blocked with 1× Tris-buffered saline–Tween 20 (TBST) containing 10% skim milk powder (Fluka) and 0.2% Tween 20 for 1 h. The primary antibody was a rabbit polyclonal antiserum recognizing phosphorylated STAT2 (Abcam; ab53132), and a secondary polyclonal antiserum was peroxidase-conjugated anti-mouse IgG(H+L) (Jackson ImmunoResearch). A chemiluminescence kit (SuperSignal West Pico Chemiluminescent Substrate; Thermo) was used in a standard Western blotting procedure following the manufacturer's recommendations.

RNA extraction.

Samples (embryos, placentae, and dam spleen) were immediately collected on ice and then blended in a 10-fold volume of TRIzol (Invitrogen) using a polytron (Kinematica). Total RNA was extracted using the manufacturer's protocol. Following RNase-free DNase I (NEB) digestion at 37°C for 30 min, the samples were then reextracted with TRIzol, precipitated, and finally resuspended in 50 μl of diethyl pyrocarbonate (DEPC)-treated RNase-free water. The RNA quality was assessed by agarose gel electrophoresis of total RNA. Samples exhibiting crisp 28S and 18S rRNA bands with intensity ratios of 2 or greater were accepted for further analysis.

cDNA synthesis.

Five micrograms of RNA was used for cDNA synthesis using oligo(dT)₁₈ as a primer (NEB) and Moloney murine leukemia virus (MMuLV) (NEB) or Superscript II (Invitrogen) reverse transcriptase and the manufacturer's (Invitrogen) protocol. Samples were diluted 1:4 with double-distilled water (ddH₂O) and stored at −20°C.

Quantitative PCR (qPCR) primer preparation.

Transcript-specific PCR primer pairs were designed in MacVector. Forward and reverse primers were in different exons, except those for IFN-β, which is a single-exon locus. The primers (Table 1) were designed and tested by gradient PCR, agarose gel electrophoresis, and melting-curve analysis. Each pair was shown to amplify the expected target only under the reaction conditions.

TABLE 1.

Primers used in the study

Locus	Primer sequence
	Forward	Reverse
GAPDH	5′ CTTCATTGACCTCAACTACATGG 3′	5′ GAGATGATGACCCTTTTGGC 3′
PKR	5′ CATAGTTGTTGGGAGGGAGTTGAC 3′	5′ TTCCATCATTTTCCAGGGCTG 3′
IFN-β	5′ GCACTGGGTGGAATGAGACTATTG 3	5′ TCTGAGGCATCAACTGACAGGTC 3′
Viral NS1	5′ ACCTTGCCTGGTGACTTTGGTTTG 3′	5′ ACTTTTCGGTGTCGTGAATGGTGAG 3′
TNF-α	5′ GCTCTTCTGTCTACTGAACTTCGG 3′	5′ ACTTGGTGGTTTGCTACGACG 3′
IL-1β	5′ AAAAAGCCTCGTGCTGCT 3′	5′ CCTGACCACTGTTGTTTCCCAG 3′
IFN-γ	5′ TCTTCAGCAACAGCAAGGCG 3′	5′ AATCTCTTCCCCACCCCGAATCAG 3′

qPCR.

Reactions (25 μl) were set up in triplicate, each containing 1 μl diluted cDNA template, 12.5 μl of PCR ready mix (Kappa Biosystems), and 0.1 μM (each) forward and reverse primers. Similar standard reactions of four 5-fold serial dilutions of purified template were set up. PCR amplification was performed on a MyIQ thermal cycler (Bio-Rad), using the following cycling protocol: initial denaturation at 95°C for 2 min and 45 cycles of denaturation at 95°C for 15 s, annealing at 56°C for 15 s, and extension at 72°C for 35 s, followed by a final extension at 72°C for 2 min. Postamplification melting-curve analyses were done to confirm reaction specificity.

In reactions where the variable of interest was the relative change in gene expression following infection, the standard method of Pfaffl (46) was used. In the case of viral burden, the relationships from which the Pfaffl formula was derived were used, and the starting fluorescent signal was calculated as follows: S₀ = SC_T/E^CT, where S₀ is the signal at start, SC_T is the signal at threshold, E is the efficiency of the reaction, and C_T is the cycle threshold. In all cases, the calculated start values were normalized to the start values of the GAPDH housekeeping gene.

RESULTS

Effect of MVMp infection in utero on embryonic transcription of IFN-β and IFN-γ.

Data from in vitro experiments showed that MVMp infection induces IFN-β in primary tissue-cultured fibroblasts (MEFs) (24, 26), and we have shown that MVMp can mount a widespread productive embryonic infection in utero, with the most frequently infected cell types being respiratory epithelia and fibroblasts (18) (Fig. 1A to C). We hypothesized that inoculating the developing mouse embryo in utero with MVMp would lead to activation of an innate immune response and induction of type I/II IFN signaling cascades. To test this, we checked embryonic transcription levels of IFN-β and IFN-γ using quantitative RT-PCR during the course of MVMp infection.

FIG 1.

MVMp infection in utero and its effect on embryonic transcription of IFN-β and IFN-γ. Embryos (13.5 days p.c.) were injected in utero with 5 × 10⁵ PFU of MVMp, background material, or sterile saline as described in Materials and Methods. Injected and control uninjected embryos from undisturbed pregnancies were collected at the indicated time points. (A to C) Representative embryonic infections. Embryos were collected, sectioned, and subjected to immunohistochemistry using an anti-MVMp capsid antiserum. The images show pairs of differential interference contrast (DIC) (top) and fluorescent (bottom) images of embryonic lung at 48 (A), 72 (B), and 96 (C) hours following inoculation. (D and E) Induction of interferons in infected embryos. cDNA prepared from isolated RNA was used in real-time quantitative-PCR assays to establish expression levels of the GAPDH housekeeping gene and the IFN-γ, IFN-β, and viral NS1 genes. Expression levels were normalized to the GAPDH housekeeping gene expression level. Average fold inductions (left axes) were calculated for replicates against values from age-matched embryos from undisturbed pregnancies (the dashed lines indicate 1-fold induction). The error bars indicate standard errors of the mean (SEM). The asterisks mark average fold induction values that are significantly different (P < 0.05 by analysis of variance [ANOVA]) from those of the untreated sample. (D) Fold induction of IFN-β and IFN-γ in infected embryos. The extent of viral infection is represented by viral NS1 gene expression signal normalized to the GAPDH housekeeping gene signal and is indicated by the diamonds and the scale on the right axis. (E) Fold induction of IFN-β and IFN-γ in uninfected embryos. IFN-β and IFN-γ levels in injected embryos (inj), contralateral uninjected embryos (cl), and embryos injected with background material (cbm) were determined 72 h after injection.

Mouse embryos (13.5 days p.c.) were inoculated with wild-type MVMp, injected with background material as a control, or untreated. After further gestation, injected embryos, contralateral uninjected littermates, and control embryos were collected and processed for either immunohistochemistry or real-time (RT)-qPCR. In all cases fold induction was calculated relative to that of age-matched embryos from unperturbed pregnancies. The data collected (Fig. 1) showed a progressive increase in the extent of infection (Fig. 1A to C) accompanied by an increasing viral burden (Fig. 1D). RT-qPCR analysis of MVMp-inoculated embryos showed that relative IFN-β transcription increased during the course of infection (Fig. 1D) and was linearly correlated with the increasing viral load (R_p = 0.862, where R_p is Pearson's coefficient of correlation; P < 0.01), finally reaching 90-fold induction at the last time point examined, 96 h after inoculation. No changes in the transcription levels of IFN-β were detected in uninoculated littermates from the contralateral uterine horn or in embryos injected with background material (Fig. 1E). Infection also significantly increased the levels of IFN-γ transcripts; however, the pattern of induction looked different. Unlike IFN-β, IFN-γ levels did not correlate with the viral burden. Infected embryos exhibited a significant 55-fold increase in IFN-γ signal at 72 h postinfection, but by 96 h, the number of transcripts dropped (Fig. 1D). Interestingly, uninjected contralateral embryos exhibited a 5-fold elevation in IFN-γ transcripts at 72 h postinoculation (Fig. 1E), but this difference failed to reach statistical significance.

Effect of MVMp infection in utero on embryonic transcription of IL-1β and TNF-α.

A typical antiviral response, at least in the adult, includes the induction of proinflammatory cytokines as a result of IFN signaling (47). We therefore looked for changes in the transcription levels of two proinflammatory cytokines, TNF-α and IL-1β, following MVMp infection using the same experimental protocol described for IFN-β and IFN-γ.

The collected data (Fig. 2) showed that there was statistically significant transcription upregulation of both TNF-α and IL-1β in infected embryos compared to controls. TNF-α transcripts were upregulated by a factor of 7 at most, and its pattern of induction (Fig. 2A) was the same as that of IFN-β (Fig. 1D), increasing throughout the course of infection analyzed. Untreated embryos contralateral to the inoculated embryos showed a slight elevation in the number of TNF-α transcripts, but compared to other untreated controls, this difference was not statistically significant (Fig. 2B).

FIG 2.

Effect of MVMp infection in utero on embryonic transcription of IL-1β and TNF-α. Injected and control embryos were collected after different incubational periods, and total RNA was extracted and used for cDNA synthesis. Real-time quantitative PCR was performed using IL-1β, TNF-α, and cellular GAPDH primers. Cytokine levels were then normalized to the corresponding cellular GAPDH levels. The representation of results is the same as that in Fig. 1D and E. (A) Fold induction of IL-1β and TNF-α in infected embryos. The extent of viral infection is represented by viral NS1 gene expression signal normalized to the GAPDH housekeeping gene signal and is indicated by diamonds and the scale on the right axis. (B) Fold induction of IL-1β and TNF-α in uninfected embryos. IL-1β and TNF-α levels in injected embryos (inj), contralateral uninjected embryos (cl), and embryos injected with background material (cbm) were determined 96 h after injection.

IL-1β transcription in infected embryos was also upregulated (Fig. 2A). There was a peak in the number of transcripts at 72 h postinoculation, an 8-fold induction, but the level dropped to half that by 96 h postinfection, a pattern of induction similar to that observed for IFN-γ. No induction of the cytokine was observed in untreated contralateral or background material-injected controls (Fig. 2B).

Effect of embryonic MVMp infection in utero on antiviral effector activity in the embryo.

At the intracellular level, the innate response triggered by viral infection includes entry into an antiviral state that abrogates the production of viral components at multiple levels. This state is mediated by the activation or induction of proteins, partially in response to IFN-stimulated genes (ISGs) that modulate basic cellular processes, including transcription and translation (reviewed in references 48 and 49). Two markers of these pathways are STAT1 and STAT2 phosphorylation and induction of the antiviral PKR gene. We hypothesized that the IFN-β-triggered antiviral pathways in the MVMp-infected embryos would also be activated.

We tested for the presence of phosphorylated STAT2 protein (130 kDa) in lysates from MVMp-infected and untreated control embryos by Western blotting (Fig. 3A). Specific bands were found in lysates collected from MVMp-infected embryos at 72 and 96 h postinoculation. No phosphorylated STAT2 protein was found in lysates from untreated controls, from uninoculated embryos contralateral to injected embryos, or from infected embryos shortly after MVMp inoculation. We used RT-qPCR to measure changes in PKR transcript levels in response to viral infection (Fig. 3B). As hypothesized, in MVMp-inoculated embryos, there was a statistically significant elevation of PKR transcripts.

FIG 3.

Effect of embryonic MVMp infection in utero on antiviral effector activity in the embryo. Embryos (13.5 days p.c.) were injected in utero with 5 × 10⁵ PFU of MVMp (inj), background material (cbm), or sterile saline (nc) as described in Materials and Methods. cl, embryos contralateral to virus-injected embryos. (A) Changes in phosphorylated STAT2 levels. Embryo lysates were prepared at the time points shown following infection, subjected to SDS-10% PAGE, and blotted. The membrane was probed with an antibody specific for phosphorylated STAT2 protein (P-STAT 2). (B) Changes in embryonic transcription of PKR. Ninety-six hours after infection, injected and control embryos were collected and total RNA was extracted and used for cDNA synthesis. Real-time quantitative PCR was performed using PKR and cellular GAPDH primers. The PKR levels were then normalized to the corresponding cellular GAPDH level. The representation of results is the same as that in Fig. 1D and E.

Cytokine production in placentae following MVMp inoculation in utero.

The data suggest that the developing mouse embryo is capable of mounting an innate immune-mediated antiviral response following MVMp inoculation in vivo. Since our infection technique minimizes maternal exposure to the virus, we hypothesized that any changes in maternal responses would be due to signaling from the embryo rather than direct maternal response to the virus. Our next step, therefore, was to check if the response we observed was restricted to embryonic tissues only. We began by testing placentae, since they have the maternal tissue closest to the infected embryo and are known to be able to mount a vigorous innate response (50). We hypothesized that if the maternal part of the placenta responds, we would see changes in cytokine transcription levels in placentae, as well as in embryos.

We analyzed placentae of injected, contralateral, and untreated controls by RT-qPCR to check for changes in the relative levels of IFN-β, IFN-γ, TNF-α, and IL-1β transcripts after different incubation periods following embryo inoculation with MVMp (Fig. 4). Individual cytokines showed minor elevations in transcript levels at 72 and 96 h following infection compared to negative controls and to the 24-h samples. None of these changes reached statistical significance. Notably, the transcript levels measured in contralateral placentae were uniformly more similar to those in the placentae of their injected littermates than to those in negative-control placentae or in contralateral placentae at other time points. The aggregate data suggest that there is a small placental response that is uniform across the set of littermates, injected and uninjected alike.

FIG 4.

Cytokine production in placentae following MVMp inoculation in utero. Selected 13.5-day p.c. embryos were injected in the liver in utero with 5 × 10⁵ PFU of canonical MVMp. Placentae from injected and uninjected embryos from the same litter were collected and analyzed at 24 h, 72 h, and 96 h. Relative induction levels were calculated as for embryo samples (Fig. 1 and 2) using real-time quantitative PCR. The representation of results is the same as that in Fig. 1D and E. The open diamonds at the base mark the bars for placental samples from uninjected contralateral embryos.

CD11b⁺ cells in embryos following MVMp infection.

The induction of IFN-β, IFN-γ, and other proinflammatory cytokines following MVMp infection of the embryo in utero is consistent with a normal mature innate immune response to viral infection. Our final goal in this study was to determine whether this would translate into the direct involvement of dedicated embryonic cells of the innate immune system. The β2 integrin CD11b is an adhesion molecule found on mature leukocytes (51), including macrophages, neutrophils, dendritic cells (DCs), and NK cells. At least in the adult, leukocytes migrate to the sites of infection in response to chemokine release (52). We hypothesized that inoculation with MVMp in vivo would lead to enhanced infiltration of these cells into infected embryonic tissues. To test this, we inoculated 13.5-day p.c. embryos with MVMp as before and let gestation proceed for another 96 h. The distributions of mature leukocytes in inoculated and control embryos were then determined by immunohistochemistry using an anti-CD11b primary antibody.

The data showed relatively high numbers of CD11b⁺ cells in the nasal mesenchyme and the liver, regardless of infection status (data not shown). In the liver, at least, this is to be expected, as it is still the major site of hematopoiesis at this developmental stage. In MVMp-infected embryos, the gut and lung showed a significant increase in the number of CD11b⁺ cells (Fig. 5C and D, right) compared to untreated or mock-injected controls (Fig. 5A and B, right). Infiltration of CD11b⁺ cells into those tissues correlated with extensive viral infection in the area (Fig. 5C and D, left). We did not observe any difference in the number or distribution of CD11b⁺ cells in the placentae of infected versus uninfected embryos.

FIG 5.

CD11b⁺ cells in embryos following MVMp infection. Embryos (13.5 days p.c.) were inoculated with 5 × 10⁵ PFU of canonical MVMp. After 96 h of incubation, injected and uninjected age-matched control embryos were collected, fixed, and sectioned as described in Materials and Methods. The distribution of MVMp-infected cells was determined by immunofluorescent staining with anti-MVMp capsid primary antibody and Cy2-coupled anti-rabbit IgG secondary antibody. The distribution of cells of the innate immune system was determined by immunofluorescent staining with anti-CD11b rat monoclonal antibody and DyLight 550-coupled anti-rat IgG secondary antibody. The image pairs show the same field, illuminating either the green MVM Cy2 signal (left) or the red CD11b DyLight550 signal (right). (A and B) Sections from uninfected embryos (−). (C and D) Sections from infected embryos (+).

DISCUSSION

The research described here had two goals. One was to establish an in vivo baseline for understanding the interaction between the mammalian parvoviruses and the innate immune system. The second was to develop a model system in which the mammalian embryonic immune system response to an embryonic viral infection could be efficiently characterized. To approach these goals, we used a murine model of direct inoculation of midtrimester embryos with MVM in utero. With respect to the first target, we reasoned that using a completely normal cell population in a model system in which we had already demonstrated a robust and widespread viral infection would increase our chances of observing an innate response. With respect to the second target, we reasoned that if we did observe a response, the lack of maternal exposure and the short incubation times would avoid complications in interpretation that might arise from an embryonic response that reflected concurrent maternal infection rather than the embryonic infection (53).

The data point to a strong and comprehensive antiviral response in the midgestation mammalian embryo. The infected mouse embryos engaged each branch of the innate immune system that we tested. Both type I (IFN-β) and type II (IFN-γ) IFN pathways were triggered. Transcription of the proinflammatory cytokines TNF-α and IL-1β increased. STAT2 and PKR, both involved in establishing an intracellular antiviral state, were upregulated. Finally, cells positive for the integrin receptor subunit CD11b, a marker for mature myeloid cells, specifically macrophages, monocytes, neutrophils, and NK cells, accumulated in MVM-infected tissues in the embryo. As far as we know, this is the first such demonstration of the recruitment of mature leukocytes to infected embryonic tissues in utero.

The results are consistent with the small body of studies pointing to a midgestation fetal innate immune response to viral infection in either animal models (40, 54) or clinical studies (55, 56). Certainly, further research is needed to better characterize the ontogeny of the mammalian innate immune system. The immunology literature contains numerous references to a comparatively “immature” embryonic innate immune system that is biased toward a tolerogenic rather than a proinflammatory response (57–59). However, this view is based largely on late-third-trimester and perinatal studies. Our data add further weight to suggestions (60) that such shifts in balance may be necessary features of the physiology of pregnancy rather than reflecting immaturity. For instance, the perinatal fetus may modulate its innate response in anticipation of the sudden exposure to the very large nonpathogenic microbial world at birth. The earlier embryo, albeit missing an adaptive response, may be able to clear viral infections via a fully functional innate response. This likely has medical consequences that are independent of the role of viral infection in fetal pathology. It is possible that an embryonic response to fetal viral infection may resolve the infection in utero, leading to apparently normal development and birth, but with long-term sequelae. If indeed a large fraction of the cases of schizophrenia and autism could be prevented by eliminating maternal influenza infection during pregnancy (61), one can reasonably speculate about the long-term health effects of a cryptic and resolved viral infection of the fetus during development.

Since our focus with respect to the host response in this research was on the embryonic response to an embryo infection, the procedure followed was designed to minimize the maternal response to the viral pathogen. Nonetheless, given the extensive communication between mother and embryo, we did expect to see some maternal response to the infected embryo. We tried to search for this response in the placenta, since the organ contains the maternal tissue that is in closest proximity to the infected embryo and is certainly capable of mounting an innate response (62, 63). Surprisingly, the data indicated a minimal placental response. The individual small elevations in innate immune factor transcription that we observed later in infection (Fig. 4) were not statistically significant. Similarly, immunohistochemically marked leukocytes did not show any change in distribution or numbers in the placenta during embryo infection. Although placental responses to individual factors usually did not reach significance, the aggregate data do indicate increased innate immune activity, but there was no difference between the responses of injected-embryo placentae versus contralateral placentae. This suggests that the placental response, minimal as it is, reflects a systemic rather than a local response to the infected embryo. This in turn reinforces the conclusion that the embryonic response to the virus is a local embryo affair; if the embryo was communicating an infected state to the dam, the signaling was not triggering a local maternal innate immune response that we were able to observe.

With respect to the parvoviruses and their potential to activate an innate immune response, we have demonstrated for the first time an unequivocal host innate immune response to the mammalian protoparvoviruses during infection in vivo. Previous studies have either failed to identify any interferon activity or induction (22, 64) or observed sporadic, low-level activity (20, 21). In vitro tissue culture studies have yielded similarly diverse results. Depending on the host and virus combination, infection may provoke an innate immune response (24, 26, 30) or not (23, 27, 30, 65). Infection may be either blocked by exogenous activation of an innate response (20, 21, 23, 24) or not (26, 27, 65–67). The experimental systems are sometimes close enough in detail that the variance in results is difficult to reconcile. The reason why the innate response we observe in utero is so much stronger than that seen in other in vivo studies is not clear. Differences in assay sensitivity could explain some of the variance. The qPCR assays that we used here may be more sensitive than the immune or bioactivity assays used in earlier studies. It is possible that the much greater extent of infection in the embryo yields a larger population of responding cells, or it may be that, in comparison to the adult, either the innate response is stronger at the level of the individual cells or the viral countermeasures repressing the response are less effective in the embryo. With respect to the latter, there is no defined protoparvoviral suppression of an innate immune response. The fact that downstream antiviral pathway elements were found to be activated indicates that, despite the reliance on transcription assays rather than bioassays, active interferon pathway proteins were produced. However, we cannot rule out partial suppression of the response by the virus. There are two observations that suggest such a mechanism may be operating. First, we observed increasing IFN-β transcription throughout infection, but IFN-γ expression and STAT-2 phosphorylation, which depend on IFN-β activity, decreased at 96 h. Second, we have no indication from this or our earlier study (18) that the embryo ever clears the infection. Indeed, by 120 h postinfection, embryos begin to die. Regardless of the relative strength of the innate immune response observed, we do not yet know whether it has any impact on the spreading viral infection in the embryo, an issue we are currently addressing.

Finally, obstetric medicine is not the only field that may benefit from a better model of autonomous-parvovirus–innate-immune interaction. Despite the close physical proximity of mice, rats, and humans over thousands of years and despite the prevalence of these parvoviruses in rodent populations, they are not zoonotic. Nevertheless, they will readily cross the species barrier to infect, replicate in, and kill human cancer cells (68, 69), but not their untransformed parental cells (70–73; reviewed in reference 74). Parvoviruses are therefore being adapted and tested for cancer therapy (75, 76) and have advanced as far as phase I clinical trials. Understanding the interactions of the oncolytic parvoviruses with the innate immune system is important in order to use them effectively in cancer therapy.