An essential role for γ-herpesvirus latency-associated nuclear antigen homolog in an acute lymphoproliferative disease of cattle

Abstract

Wildebeests carry asymptomatically alcelaphine herpesvirus 1 (AlHV-1), a γ-herpesvirus inducing malignant catarrhal fever (MCF) to several ruminant species (including cattle). This acute and lethal lymphoproliferative disease occurs after a prolonged asymptomatic incubation period after transmission. Our recent findings with the rabbit model indicated that AlHV-1 infection is not productive during MCF. Here, we investigated whether latency establishment could explain this apparent absence of productive infection and sought to determine its role in MCF pathogenesis. First, whole-genome cellular and viral gene expression analyses were performed in lymph nodes of MCF-developing calves. Whereas a severe disruption in cellular genes was observed, only 10% of the entire AlHV-1 genome was expressed, contrasting with the 45% observed during productive infection in vitro. In vivo, the expressed viral genes included the latency-associated nuclear antigen homolog ORF73 but none of the regions known to be essential for productive infection. Next, genomic conformation analyses revealed that AlHV-1 was essentially episomal, further suggesting that MCF might be the consequence of a latent infection rather than abortive lytic infection. This hypothesis was further supported by the high frequencies of infected CD8⁺ T cells during MCF using immunodetection of ORF73 protein and single-cell RT-PCR approaches. Finally, the role of latency-associated ORF73 was addressed. A lack of ORF73 did not impair initial virus replication in vivo, but it rendered AlHV-1 unable to induce MCF and persist in vivo and conferred protection against a lethal challenge with a WT virus. Together, these findings suggest that a latent infection is essential for MCF induction.

About 1.3 million wildebeests (Connochaetes taurinus) migrate over 1,800 miles in eastern Africa every year (1, 2). Local farmers, like the Maasai people, are nomadic pastoralists who center their lifestyle on livestock herding, mostly in grazing areas shared with wildlife. The great annual spectacle of the wildebeest migration directly endangers the livelihood of the local herders, because cattle are highly susceptible to a fatal viral disease named malignant catarrhal fever (MCF), which is mainly transmitted by the 500,000 wildebeest calves born each year.

MCF is an acute, sporadic, and fatal pansystemic lymphoproliferative disease of a variety of species of the Artiodactyla order, including cattle. The main causative agents of MCF are two γ-herpesviruses that have been recently grouped in the Macavirus genus, ovine herpesvirus 2 (OvHV-2) and alcelaphine herpesvirus 1 (AlHV-1). These viruses cause no apparent disease in their natural host species. Sheep are naturally infected by OvHV-2, which is responsible for the sporadic sheep-associated form of MCF. Wildebeests carry AlHV-1, responsible for the wildebeest-derived form of the disease (3, 4). The prevalence of AlHV-1 infection in wildebeest is close to 100%, and transmission mainly occurs during the calving period and in the first 3–4 mo of life (5, 6). MCF impact on the local pastoralist populations has largely been underestimated, with recent reports showing that MCF is perceived to be the cattle disease with the highest economical and social impacts in these areas (7, 8). In addition, MCF has been reported throughout the world in game farms or zoological collections where mixed ruminant species, including wildebeest, are kept (9).

The mechanisms responsible for the lymphoproliferative and degenerative lesions observed in MCF are unknown (3, 10, 11). Initially, the very low levels of detection of infected cells in lesions led to the hypothesis that MCF could be caused by very few infected cells interacting with the surrounding uninfected T cells, resulting in their deregulation (12, 13). However, recent reports have suggested that virus infection in vivo might be more frequent than previously thought (14, 15). MCF can be experimentally induced in rabbits, where the observed lesions are indistinguishable from the lesions described in the MCF-susceptible species (16). Using this model, we have recently shown that AlHV-1 infection is responsible for the induction of a severe proliferation of CD8⁺ T cells in peripheral blood mononuclear cells (PBMCs) and lymphoid organs (14). It has also been shown that the infection is restricted to CD8⁺ cells in PBMCs and that at least 10% of these cells in PBMCs contain the viral genome. Other than the proliferation of CD8⁺ T cells in lymphoid tissues, MCF is characterized by the infiltration of activated and cytotoxic CD3⁺CD8⁺CD4⁻ T cells in the perivascular spaces of all tissues and organs (17). Using a recombinant virus strain of AlHV-1 expressing the firefly luciferase, we recently showed that the macroscopic distribution of AlHV-1 infection in explanted organs of MCF-developing rabbits colocalizes with the distribution of lesions in both lymphoid and nonlymphoid tissues of MCF-developing rabbits (15). The lack of detection of infectious viral particles together with the low (or no) expression of few selected viral genes normally expressed during productive viral infection suggested the absence or rarity of cells supporting productive infection in the tissues (14, 15). These results suggested that AlHV-1 infection might not be productive during MCF and could be latent.

Here, we investigated whether latency establishment could explain the apparent absence of productive infection and examined the role of AlHV-1 latency in MCF pathogenesis. First, we studied MCF in calves after experimental AlHV-1 infection and used a whole-genome approach to analyze both cellular and viral RNA expression in the lymph nodes of MCF-developing calves. Our findings strongly confirmed an absence of productive infection in MCF together with a profound disruption of the cellular gene expression profile. Next, we examined the viral genomic conformation and revealed that AlHV-1 genomes are essentially maintained as latent episomes in the tissues of MCF-developing calves and rabbits, a signature of classical γ-herpesvirus latency. Our hypothesis suggesting that MCF might be a consequence of latency was further confirmed by the high frequency of infected CD8⁺ T cells in the cellular infiltrates of MCF-developing animals. Persistence in actively dividing cells during γ-herpesvirus latency is dependent on the expression of the latency-associated nuclear antigen (LANA) homolog, also termed genome maintenance protein, encoded by ORF73 (18). Thus, we addressed the role of AlHV-1 latency in MCF by producing ORF73-deficient viruses. Whereas ORF73 disruption did not impair viral replication in vitro and in vivo at the first days postinfection (p.i.), a lack of ORF73 rendered AlHV-1 unable to induce MCF and persist in vivo. Of importance, we finally showed that infection with an ORF73-deleted recombinant virus conferred protection against a lethal challenge with a WT virus. Altogether, our findings suggest that AlHV-1 ORF73-mediated latency is a prerequisite for the induction of the lymphoproliferative lesions observed in MCF and bring prospects on future vaccine development for populations living in areas where AlHV-1 infection is endemic.

Results

Design of a Combined Array to Detect Viral and Cellular Gene Expression.

We designed an oligonucleotide microarray combining 17,408 tiling 60-mer probes complementary to each strand of the AlHV-1 genome and 43,603 probes of the Bovine Gene Expression Microarray (Agilent Technologies). In this design, more than 70% of the array consisted in cellular probes. After having verified that more than 75% of the overall probes were not differentially expressed, a classical norm exp background correction followed by loess normalization was applied. To visualize the viral tiling array data, we averaged the log₂ ratio fluorescence intensities for every probe overlapping each nucleotide in the AlHV-1 genome. To analyze differentially expressed cellular probes, a linear model with a dye-swap blocking variable was fit to the data using the least squares method. An empirical Bayes-moderated t statistic was then used to test the AlHV-1/Mock contrast, and a Benjamini–Hochberg multiple testing correction was applied.

Detection of Viral Gene Expression During Lytic Infection.

We first examined the expression of viral polyadenylated RNA from infected [multiplicity of infection (moi) = 0.01] madin-darby bovine kidney (MDBK) cells at 72 h p.i.. This late time point was chosen, because we expect the majority of viral genes to be expressed. The expression of viral RNA at each nucleotide position obtained from two-color dye-swap analyses of four independent biological repeats is shown in Fig. 1A compared with the annotated ORF of the AlHV-1 genome. Overall, RNA expression was observed on 45% of the total length of both strands of the AlHV-1 genome. The results of each individual repeat are plotted in SI Appendix, Fig. S2 and show very little signal variation, indicating high reproducibility. The characteristic progressive loss of signal from the 3′ to the 5′ end of most transcripts can be explained by the use of oligo-dT primers in the amplification step before hybridization. Expressed viral transcripts were, therefore, identified using the wavelet transform as a multiscale peak detector on the averaged viral signal of each strand. Briefly, peaks with a nearby polyadenylation signal were defined as 3′ ends, and corresponding 5′ ends were defined manually. Most signal peaks corresponded to previously annotated ORF boundaries, which indicated that the designed tiling array detected known expressed transcripts. Among all annotated ORFs, only three ORFs were not associated with an expressed transcript (A5, ORF18, and ORF56). In addition to predicted ORFs, 23 additional expressed regions (ERs) did not correspond to annotated ORFs (Fig. 1A and SI Appendix, Table S1). Their roles during viral infection are unknown. As a whole, these results not only experimentally confirmed 96% of viral annotated ORFs and identified 23 additional ER, but they also validated our tiling array design in the detection of viral gene expression during AlHV-1 infection.

Fig. 1.

Genome-wide viral RNA expression during AlHV-1 infection in vitro and during MCF. (A) MDBK cells were infected with AlHV-1 at moi = 0.01, and total RNA was extracted 72 h p.i., labeled, and hybridized to the custom-designed array (n = 4). (B) Inguinal LNs were harvested from mock- or AlHV-1–infected calves developing MCF (n = 4). Total RNA was extracted, labeled, and hybridized to the array. The mean fluorescence ratio of Cy5 and Cy3 dyes between AlHV-1– and mock-infected samples of probes overlapping each nucleotide position is plotted on a log₂ scale for each strand. The red and blue curves show the forward and reverse strands of the genome, respectively. Red and blue block arrows represent annotated genes on the forward and reverse strands of the genome, respectively. Peaks identified by the wavelet transform close to a poly-A signal (AATAAA or ATTAAA) are indicated with dotted lines. Expressed regions are shown with gray boxes (dark gray, peak detection with poly-A signal and overlapping with a predicted ORF; medium gray, peak detection with poly-A signal but no overlapping with a predicted ORF; light gray, peak detection with no poly-A signal). E, expressed region not overlapping an annotated ORF.

Cellular and Viral Gene Expression Profiles During MCF.

We experimentally infected calves with AlHV-1 to induce MCF. All AlHV-1–infected calves developed typical prostration, nasal discharge, lymphadenopathy, and severe persistent hyperthermia at 12.2 ± 0.5 d p.i. (SI Appendix, Fig. S1A). All AlHV-1–infected animals developed typical MCF with lymph node (LN) hypertrophy, increased CD8⁺ T-cell percentages in peripheral blood and LN, increased IFN-γ production, and characteristic infiltrations of lymphoblastoid cells in the perivascular spaces of many tissues (SI Appendix, Fig. S1).

To determine cellular and viral gene expression during MCF, we extracted RNA from the inguinal LN of each calf for analysis on a custom-designed array as described above. The choice of inguinal LN as the selected tissue was arbitrary and based on the fact that AlHV-1 viral genomic load is the highest in peripheral LN. Cellular and viral RNA transcription profiles were analyzed identically to viral gene expression during lytic infection in vitro (two-color dye-swap analyses of four independent biological repeats).

We observed in the LN of MCF-developing calves a total of 225 differentially expressed (P value ≤ 10⁻⁴, fold change was >4 or <−4) genes, from which 55 were overexpressed and 170 were underexpressed during MCF (SI Appendix, Fig. S4 and Table S2). Among the highest overexpressed genes, we detected the cytotoxic granule proteases granulysin and granzymes, proinflammatory secreted proteins [including IFN-γ, TNF-α, chemokine (C-C motif) ligand (CCL)-3, CCL-4, and chemokine (C-X-C motif) ligand (CXCL)-11], and proteins involved in cell division and proliferation (including DNA helicase Pif1, cyclin A2 and B1, CDC28 protein kinase regulatory subunit 2, and cell division cycle associated 2). Among underexpressed proteins, we found many receptors and transmembrane proteins, proteins involved in cellular adhesion, and three genes that have been involved in tumor suppression (Dab2, Armcx1, and Reck). Overall, these data show a severe disruption of host gene expression in LN tissue during MCF.

We next analyzed the AlHV-1 genome tiling array to determine the viral gene expression in LN during MCF. The mean log₂ ratio intensities per nucleotide are shown in Fig. 1B. The results of each individual repeat in SI Appendix, Fig. S3 show that the analyses were highly reproducible, with nearly all nucleotide positions showing very little variation between the four biological replicates. Compared with productive infection (Fig. 1A), viral RNA expression during MCF was very sparse throughout the AlHV-1 genome, with only 10% of each strand being expressed. The expression levels of some transcripts (i.e., ORF52, ER13, and ORF73) were similar to the signal intensities observed during productive infection, suggesting that the observed scarce viral gene expression is not caused by low sensitivity. We detected the expression of only 16 annotated ORFs (ORF17, -23, -45, -46, -A6, -52, -53, -55, -58, -59, -60, -61, -62, -65, -69, and -73) and 8 ERs that were not predicted as an ORF (ER1, -11, -12, -13, -14, -15, -16, and -19) (Fig. 1B and SI Appendix, Table S1). Apart from the high expression of ORF73, which encodes the LANA homolog or genome maintenance protein (18), the role in γ-herpesvirus infection of most of these genes remains largely unknown. Moreover, the expression levels of genes known to be essential and/or highly expressed during productive infection, such as ORF8 (gB), ORF9 (DNA polymerase), ORF22 (gH), ORF25 and -26 (capsid proteins), and ORF50 (reactivation transactivator), were under the detection level. These observations, therefore, strongly suggest that AlHV-1 infection is not productive during MCF.

Episomal Maintenance of AlHV-1 Genomes During MCF.

Whereas linear viral genomes are found in cells supporting a productive infection, herpesvirus latency is characterized by the maintenance of covalently closed viral episomes (18). We, therefore, used Gardella gel electrophoresis to analyze the virus genome conformation in MCF. AlHV-1 genomes were almost exclusively circular in PBMCs and LNs of MCF-developing calves and rabbits, respectively (Fig. 2A). Episomal AlHV-1 genomes were also predominant in PBMCs, LNs, spleen cells, and leukocytic cells from liver and kidney of MCF-developing rabbits (Fig. 2 A and B). Serial dilutions of BAC plasmid DNA and rabbit LN cells followed by Gardella gel electrophoresis gave an estimation of ∼2–5 × 10⁵ episomes per 10⁵ cells (Fig. 2C). This calculation was based on the similar signal intensities observed for 1.6 and 8 ng BAC plasmid DNA and 0.45 and 0.9 × 10⁶ cells, respectively (1 ng 150 kbp AlHV-1 genome ≅ 6.2 × 10⁶ copies). A similar analysis performed on bovine lymphocytic cell lines (LCLs) propagated from MCF-developing calves also showed episomal persistence in these lines (19). These results support classical γ-herpesvirus latency during MCF.

Fig. 2.

Episomal maintenance of AlHV-1 genomes during MCF. (A) Gardella gel analysis of AlHV-1 genome conformation during MCF in calf peripheral blood or rabbit popliteal LN cells (lanes 1 and 2 show the results obtained from two different animals). Cells were lysed in situ, and extracts were electrophoresed through a 0.8% agarose-Tris-Borate EDTA gel before Southern blotting performed with AlHV-1–specific probes. BAC plasmid DNA was used as positive control to detect covalently closed circular (CCC) viral DNA. AlHV-1–infected MDBK cells supporting a lytic replication [±90% cytopathic effect (CPE)] were used to detect linear virus genome. (B) Analysis of AlHV-1 genome conformation during MCF in rabbit lymphoid and nonlymphoid tissues as in A. (C) Gardella gel analysis of MCF-developing rabbit popliteal LNs. Fivefold dilutions of BAC plasmid DNA control and twofold decreasing cell numbers are shown.

Infiltration of ORF73-Expressing T Cells in MCF Lesions.

Latent infection of proliferating cells by γ-herpesviruses is associated with the expression of the genome maintenance protein encoded by ORF73 in AlHV-1 (20). To examine ORF73 expression in lesions, we produced antibodies specific to AlHV-1 ORF73 C-terminal domain by DNA immunization (SI Appendix, SI Materials and Methods). Specific intranuclear staining was verified after transient cell transfection with p73-V5 and on infected cells in vitro (SI Appendix, Fig. S5 A and B). Tissue cryosections were obtained from MCF-developing calves or rabbits after AlHV-1 infection. Staining specificity was controlled using tissue from mock-infected animals or naive mouse serum as primary antibody on tissue from MCF-developing animals (Fig. 3 A, Left and B, Left). Specific intranuclear staining was only observed in MCF-developing animal tissues using anti-73C polyserum. We observed high expression of ORF73 in LNs of calves and rabbits (Fig. 3 A and B and SI Appendix, Fig. S5 C and D). The experimental rabbit model was then used to investigate ORF73 expression in cells infiltrating the tissues. We observed that a high number of infiltrating T cells expressed the ORF73 protein (Fig. 3 B–E). The percentages of ORF73-expressing cells in the cellular infiltrates were estimated to be around 20% in liver and lung infiltrates (Fig. 3E). Some infected cells might express low levels of ORF73 that could be undetectable by immunostaining. Therefore, a single-cell RT-PCR approach was conducted on cloned rabbit LN CD8⁺ T cells from mock-infected or MCF-developing rabbits (SI Appendix, Fig. S6). This approach enables amplification of both viral ORF73 cDNA and viral genomic DNA to increase the chances of detection. We detected as much as 43% (10/23) to 78% (18/23) of infected CD8⁺ T cells in the popliteal LN. Together with previous observations, these results strongly suggest that the lesions occurring during MCF can be attributed to the proliferation and infiltration of latently infected T cells.

Fig. 3.

In situ expression of AlHV-1 ORF73 during MCF. (A) Immunostaining of ORF73 expression in cryosections of calf axillary LNs and rabbit popliteal LNs in mock-infected animals (mock) or during MCF (AlHV-1). White arrows show positive cells with intranuclear ORF73 staining. (B) Immunostaining of ORF73 expression in rabbit popliteal LNs, spleen, lung, liver, and kidney tissue during MCF. Microphotographs are centered on infiltrative lesions. (Magnification: 200×.) (C) Spleen cryosection showing intranuclear ORF73 staining. White arrows show positive cells with intranuclear ORF73 staining. (D) ORF73 and CD3 costaining of rabbit spleen and kidney cryosections. Immunostainings were performed with homemade anti-73C mouse polyserum. Stainings were revealed with Alexa-Fluor 488 nm goat anti-mouse IgG secondary antibody (AF488) and DAPI used for counterstaining. (Magnification: 100×.) (E) Percentage of ORF73-positive cells in lung (n = 10) and liver (n = 12) cellular infiltrates of two MCF-developing rabbits. Infiltrative areas were photographed and delimited before the number of DAPI⁺ nuclei was counted using ImageJ software and the cell counter plugin followed by the determination of ORF73⁺ cell numbers.

AlHV-1 ORF73 Is Not Essential for Viral Replication but Is Essential for the Induction of MCF.

The role of AlHV-1 latency in the pathogenesis of MCF was subsequently addressed by producing ORF73-deleted (73^null) and -nonsense (73^ns) recombinant viruses together with their respective revertant strains (Fig. 4 and SI Appendix, Fig. S9). The lack of ORF73 did not affect protein expression levels of the envelope gp115 complex in infected cells (SI Appendix, Fig. S10 A and B) or viral growth in vitro (SI Appendix, Fig. S10C). Absence of virus growth defect in culture was in line with some reports showing that γ-herpesvirus ORF73 is not essential for virus replication in vitro (21–24), although MuHV-4 or Rhesus monkey rhadinovirus showed lytic replication deficit under certain conditions or enhancement, respectively (25, 26). To further address the effect of ORF73 deletion on virus replication in vivo, we used the luciferase⁺ WT AlHV-1 (247N-luc⁺) BAC plasmid (15) to produce a 247N-luc⁺-73^null recombinant virus (SI Appendix, Fig. S11). We then infected rabbits intranasally and monitored luciferase signals by ex vivo bioluminescence imaging in the nasal turbinates and the lungs (SI Appendix, Fig. S12). We observed that disruption of ORF73 did not impair host colonization at days 2 and 4 p.i., suggesting that ORF73 is not essential for virus replication in vivo.

Fig. 4.

Generation of the ORF73-deleted (73^null) virus strain. (A) Recombineering methodology used to delete the entire ORF73 coding sequence by galK insertion and generation of the 73^null and 73-Rev strains. (B) Flowchart of stages performed to produce the 73^null and 73-Rev strains. (C) Southern blotting analysis of produced BAC plasmids and reconstituted strain genomic DNA after HindIII restriction. The probes used for Southern blotting are indicated: C500, entire BAC WT plasmid DNA digested with SacI; ORF73, PCR amplicon corresponding to nucleotides 116,438–116,681 of the AlHV-1 genome; galK, entire galK coding sequence. (D) Confocal microscopy of MDBK infected with the BAC WT, 73^null, and 73-Rev strains and stained with anti-73C polyserum as primary antibody and Alexa 568 nm-conjugated goat anti-mouse IgG (AF568) as secondary antibody. Nuclei were counterstained with TO-PRO-3 iodide (TOPRO3).

We then used the rabbit model to induce MCF. After infection, the development of MCF was monitored at regular intervals with particular attention to body temperature and hypertrophy of the popliteal LN. Rabbits infected with the 73^null or 73^ns recombinant strain did not develop MCF-typical hyperthermia or lesions and survived the infection (Fig. 5 and SI Appendix, Fig. S13 A–C), whereas control groups developed hyperthermia, LN and spleen hypertrophy, and MCF-typical histopathological lesions. CD8⁺ T-cell expansion and proliferation were further monitored by flow cytometry analyses and in vivo BrdU incorporation. Rabbits infected with the 73^null or 73^ns recombinant strain showed no CD8⁺ T-cell expansion or proliferation (SI Appendix, Figs. S7 A–C and S13D). Together, these results show that the lack of ORF73 renders AlHV-1 unable to induce MCF.

Fig. 5.

ORF73 deletion renders AlHV-1 unable to induce MCF. (A) Body temperature was recorded daily after rabbit infection with the WT, 73^null, or 73-Rev recombinant viruses. Mock-infected rabbits received uninfected BT cells. Rabbits were identified according to the day of euthanasia p.i. (see numbers). (B) Cumulative incidence of survival of the four groups of four rabbits infected i.v. with mock-infected BT cells or BT cells infected with the WT, 73^null, or 73-Rev virus strains. (C) Spleen and popliteal LN weight at the time of euthanasia. Bars represent mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001 (one-way ANOVA with Bonferroni posttest). (D) Histopathological characterization of MCF lesions observed in kidney, liver, and lung of one rabbit representative of each group. White arrowheads indicate typical infiltrations of lymphoblastoid cells. A, arteriole; Bi, small bile duct; Br, bronchi; Hp, hepatocyte; RC, renal corpuscule; T, uriniferous tubule; V, vein. Equivalent results were obtained in two independent experiments.

AlHV-1 Persistence in Vivo Depends on ORF73 Expression.

ORF73 orthologs have been shown to be critical for maintenance of the viral genome in latently infected cells (18). To measure the viral load in vivo after rabbit infection, viral genome copy numbers were determined in lymphoid tissue by quantitative PCR (qPCR) during MCF (Fig. 6 and SI Appendix, Fig. S13E). Although AlHV-1 viral load increased exponentially in PBMCs a few days before death in rabbits infected with the WT and revertant strains, no virus could be detected after infection with the 73^null strain at any time point after infection (Fig. 6A). At time of euthanasia, comparable virus reactivation was observed in cocultures of LN cells isolated from rabbits infected with WT or 73-Rev strains, but no infectious center could be detected in LN cells from rabbits infected with the 73^null strain (Fig. 6B). This result was not caused by a reactivation defect, because no viral DNA was detected after infection with the 73^null or 73^ns strains in PBMCs, LNs, and spleen cells; also, the expected viral loads were observed in animals infected with the control strains (Fig. 6C and SI Appendix, Fig. S13E). These results show that the deletion of AlHV-1 ORF73 impairs viral persistence in MCF-susceptible rabbits.

Fig. 6.

Loss of viral persistence in absence of ORF73. (A) Viral load was assayed by qPCR of viral genome copies in PBMCs over time after infection of rabbits with the WT, 73^null, or 73-Rev recombinant strains. Real-time PCR quantification was normalized on β-globin cellular genomic sequence. Data are plotted as individual measurements (n = 4). (B) Infectivity from popliteal LN cells was tested by infectious center assay of single-cell suspensions on MDBK-overlaid media containing 0.6% carboxymethylcellulose to obtain isolated syncytia as described in SI Appendix, SI Materials and Methods. Data are plotted as means ± SEM (n = 4). (C) qPCR of viral genome copies in popliteal LN at the time of euthanasia of rabbits infected with the WT, 73^null, or 73-Rev recombinant strains. Real-time PCR quantification was normalized on β-globin cellular genomic sequence. Data are plotted as means ± SD of triplicate measurements for each sample (n = 4). Equivalent results were obtained in two independent experiments.

Antiviral Humoral Response in Absence of ORF73.

Although the lack of ORF73 expression renders AlHV-1 unable to persist in vivo, the 73^null strain should induce an antiviral response. Thus, we examined the antiviral antibody response after infection with the 73^null strain. Specific anti–AlHV-1 antibodies could be detected as soon as 6–10 d after infection and reached a peak of production just before death in rabbits infected with the WT or 73-Rev strain (SI Appendix, Fig. S8). Animals infected with the 73^null virus also developed an anti–AlHV-1 antibody response. This response was, however, reduced at the late time points p.i. compared with the WT and 73-Rev groups. These results show that the deletion of ORF73 does not impair the development of antiviral antibody response.

Infection with 73^null Virus Protects Against a Challenge with Pathogenic WT AlHV-1.

Induction of an antiviral response after infection with the apathogenic 73^null strain could protect against another challenge with the pathogenic WT strain. To test this hypothesis, we infected rabbits i.v. with the 73^null strain before a challenge 28 d later with the WT virus by intranasal inoculation to mimic the natural infection route (Fig. 7A). Control groups received a mock-infected inoculum. The development of MCF was monitored daily for hyperthermia and hypertrophy of popliteal LNs. All rabbits preinfected with the 73^null strain survived the challenge, whereas four of five mock-infected rabbits developed MCF after the challenge (Fig. 7B). Antibody ELISA showed an effective anti–AlHV-1 antibody production in rabbits preinfected with the 73^null strain (Fig. 7C). Flow cytometry analyses of PBMCs revealed that rabbits preinfected with the 73^null strain did not show CD8⁺ T-cell proliferation, whereas all rabbits developing MCF in the control group had highly increased CD8⁺ T-cell percentages (Fig. 7D). Finally, rabbits showing protection against challenge did not develop typical perivascular infiltration of lymphoblastoid cells in the tissues as opposed to the unprotected control group (Fig. 7E).

Fig. 7.

73^null Infection induces protection against MCF after challenge with WT AlHV-1. (A) Experimental procedure. Three groups were used, each consisting of five rabbits. The 73^null/WT group received 10⁵ pfu 73^null strain i.v., whereas the two remaining mock/WT and mock/mock groups received uninfected vehicle BT cells. At 28 d p.i., rabbits of the 73^null/WT and mock/WT groups were challenged intranasally with 10⁵ pfu WT pathogenic strain. The mock/mock group received PBS intranasally. (B) Cumulative incidence of survival of the rabbits of each group. (C) Antibody indirect ELISA for detection of anti–AlHV-1 antibodies in serum of rabbits throughout the experiment. (D) Percentages of CD8⁺ cells in the gated T-cell population analyzed by flow cytometry at regular intervals throughout the experiment. (E) Histopathological characterization of MCF lesions observed in kidney, liver, and lung of one rabbit representative of each group. White arrowheads indicate typical infiltrations of lymphoblastoid cells. Av, arteriole; Bi, small bile duct; Hp, hepatocyte; RC, renal corpuscule; T, uriniferous tubule; V, vein. Equivalent results were obtained in two independent experiments.

Discussion

Herpesviruses coevolve with their natural host species, which often results in the virus adaptation to its host. AlHV-1 is a good example of coevolution, where the virus is so adapted to its natural host that it is able to persist without causing any clinical sign or lesion, resulting in the infection of virtually the entire wildebeest population. Contrastingly, AlHV-1 transmission to other ruminant species (e.g., cattle) causes the development of MCF that, ultimately, leads to the death of the affected animal. This relationship between AlHV-1 and wildebeest could be seen as only beneficial for the virus. However, there is also the possibility that this relationship results from an evolutionary symbiotic adaptation for both species. Under this hypothesis, AlHV-1 would take advantage of its adaptation to its host to persist, and wildebeests would benefit from MCF induction in competing species, because it would provide access to more food and/or weakened sick animals for large predators during the wildebeest calving period, corresponding to the period when virus transmission mostly occurs. Understanding how AlHV-1 is able to induce MCF to susceptible species like cattle is an important step to the understanding of this unique evolutionary relationship between a virus and its hosts.

Despite various theories, mostly limited to the description of the characteristic infiltrative lesions, no clear view currently exists to explain how AlHV-1 infection leads to disease development in MCF-susceptible species. Our recent observations in the rabbit model suggested that AlHV-1 might be predominantly not productive during MCF. In this study, we first sought to extend our findings obtained with the experimental rabbit model to species developing MCF in natural conditions. We, therefore, studied MCF experimental induction after AlHV-1 infection of calves. We designed a high-density tiling array of the entire AlHV-1 genome to examine polyadenylated viral RNA expression in vitro and during MCF in calves. This kind of approach has been used before in various herpesvirus infections (27–29), and a recent report on MuHV-4 infection has further shown the robustness of whole-genome tiling array analyses (29). Of the 90 genomic regions expressed during productive infection in cell culture, our analyses identified only 16 annotated ORFs and 8 unpredicted regions that were transcribed in LNs during MCF. Interestingly, nearly all annotated ORFs were expressed during productive infection in vitro, and we detected the expression of 23 additional ERs that were not assigned to an ORF. Non-ORF ERs were also detected in other γ-herpesvirus infection (29, 30), which redefines the genetics of viral RNA expression during AlHV-1 infection and the need of future experiments to address the role of these additional ERs during AlHV-1 infection. Consistent with our previous observations, ORF73 was expressed in MCF, whereas no expression of genes essential for virus productive infection was detected. In addition, we observed the expression of additional viral genes with mostly unknown function in infection (Fig. 1 and SI Appendix, Tables S1 and S4). It remains to be determined whether these genes and ERs are important in the pathogenesis. Interestingly, a recent report on OvHV-2–caused MCF in cattle has identified the sole expression of ORF73 and an unpredicted region corresponding to AlHV-1 ER16 (31). Although no function could be attributed to this region, it might be involved in the pathogenesis of MCF.

Our previous results in the rabbit model showed the low or absence of productive virus infection in MCF (14, 15), and we confirmed and extended these results in a whole-genome approach in calves. However, it remained unclear whether AlHV-1 persistence was mediated by classical episomal maintenance in proliferating cells or rather, explained, for instance, by a succession of abortive lytic cycles. We observed by Gardella gel electrophoresis performed directly on lymphocytic cells isolated from MCF-developing calves and rabbits that AlHV-1 genomes were essentially circular in both species and also in cells infiltrating nonlymphoid tissues such as liver and kidney, strongly suggesting classical γ-herpesvirus latency in MCF with episomal persistence of AlHV-1 genomes. This result is important, because it further validates the rabbit experimental model and also supports the hypothesis of a proliferation of latently infected cells being responsible for MCF (14, 15). A causative role of latently infected proliferating CD8⁺ T cells implies that a high percentage of T cells infiltrating the perivascular spaces in MCF are infected. Infected cells in situ were detected using homemade anti-ORF73 polyclonal antibodies, showing that a high frequency of T cells infiltrating the tissues of lymphoid and nonlymphoid organs expressed the latency-associated protein (Fig. 3). Moreover, we also showed using a single-cell RT-PCR approach that a very high percentage of CD8⁺ T cells (43–78%) (SI Appendix, Fig. S6) in the LNs of MCF-developing rabbits is infected by AlHV-1. Based on these results and together with previous data, we can conclude that latency is most probably a prerequisite for the induction of MCF. This hypothesis is consistent with latency-mediated malignancies observed in some human γ-herpesvirus infections, such as EBV or Kaposi sarcoma-associated herpesvirus (KSHV) (32, 33).

To determine the role of AlHV-1 latency-associated ORF73 in MCF, we used the 73^null and 73^ns recombinant viruses, both impaired for the expression of ORF73. Growth kinetics analyses in permissive bovine nasal fibroblasts and testis primary cell cultures did not reveal any significant deficit of the AlHV-1 ORF73-deficient viruses compared with WT and the revertant strains, suggesting that ORF73 is not necessary for virus replication. γHV-68 ORF73-deficient virus displayed substantial growth deficit in murine fibroblasts at very low moi (25), and ORF73 disruption in Rhesus monkey rhadinovirus generated a highly lytic virus strain (26). Other than these two examples, most of the ORF73-deficient γ-herpesviruses tested so far have not shown any growth defect in vitro (21–24, 34). These reports suggest that ORF73 might affect viral growth in cell culture only under certain specific conditions. Very few in vivo animal models are readily available to test viral replication at the primary site of infection in vivo. Intranasal infection of mice with MuHV-4 ORF73-deficient viruses revealed similar lung infectious titers (22) or attenuation in the acute lung infection (23, 34). In an attempt to clarify whether viral replication occurs at the primary site of infection in absence of ORF73 expression, we produced and used a 247Nluc⁺-73^null recombinant virus to infect rabbits intranasally. This approach enabled the detection of virus host colonization as soon as 2 d p.i. in the lung and the nasal turbinates (SI Appendix, Fig. S12). Ex vivo bioluminescence analyses showed that the disruption of ORF73 in AlHV-1 did not impair host colonization at the early time points p.i. and even led to increased signal intensities in the nose and lungs of some rabbits. Infection with a MuHV-4 ORF73-deficient virus resulted in replication that was concentrated to the inoculum site, suggesting that MuHV-4 ORF73 could be important for effective host colonization (35). To date, host colonization of AlHV-1 after primary infection is poorly understood, and future experiments are needed to elucidate how AlHV-1 invades MCF-susceptible hosts. Nonetheless, we provide in this study good indications that a lack of ORF73 does not affect viral replication in vivo, at least up to 4 d p.i.

Monitoring of MCF induction in rabbits after infection with 73^null and 73^ns recombinant viruses showed that a lack of ORF73 expression in AlHV-1 infection of rabbits resulted in their survival, whereas all rabbits infected with WT or revertant strains succumbed to the infection with typical clinical signs and lesions. Rabbits infected with 73^null and 73^ns virus strains did not show hyperthermia or hypertrophy of spleen and LNs and did not develop typical perivascular infiltrations of lymphoblastoid cells in kidney, liver, or lung. Moreover, we did not observe any modification of the percentages of proliferating CD8⁺ cells using in vivo BrdU incorporation after infection with the ORF73-deficient strains, whereas CD8⁺ cells severely proliferated after infection with the WT or revertant strains as observed before (14). The five strains used in this study were produced with a limited number of passages in cell culture (less than five) and showed unaltered molecular genomic structure and unaffected replication capability. High passage in cell culture has, indeed, been shown to produce apathogenic AlHV-1 strains (36–38). There is, however, no evidence of any implication of ORF73 in the loss of virulence because high passage. The causative role of ORF73 in MCF was observed in independent experiments using normalized infectious doses between each strain and two independently engineered mutant strains. Also, some of the rabbits were kept up to 6 mo after infection and did not show any MCF clinical sign or lesion, further supporting that the deletion of ORF73 renders AlHV-1 apathogenic.

The genome maintenance protein encoded by ORF73 in γ-herpesvirus tethers the viral episome to cellular chromosomes during cell division and therefore, is essential for virus persistence in vivo (22–24). One of the theories to explain MCF induction was that AlHV-1 infection could create an autoimmune-like environment, resulting in the trans activation and proliferation of uninfected T cells. After activation, these uninfected cells would not need any more presence of virus to amplify and cause MCF. However, MCF induction is associated with an exponential increasing of viral genomes at the later time points p.i., whereas no virus genome copy could be detected in PBMCs or lymphoid tissue after infection with the apathogenic ORF73-deficient virus strains. Given that ORF73 is not essential for viral replication in vitro or in vivo at early time points p.i., we can speculate that ORF73-mediated AlHV-1 episomal maintenance in vivo represents an essential prerequisite for the development of MCF. As a consequence, proliferation of CD8⁺ T cells in MCF would be caused by cis-acting mechanisms. Whether ORF73 is solely necessary for virus persistence in vivo or has additional pathogenic function to induce the observed acute lymphoproliferative lesions is unknown and will need additional investigations. Although the main function of KSHV ORF73 is to maintain the viral episome (39), it has been shown that LANA can also directly interfere with important antitumorigenic pathways, such as, for examples, the inhibition of p53-mediated functions (40, 41), RB–E2F tumor suppressor pathways (42), and antiproliferative TGF-β signaling (43). KSHV LANA is also able to up-regulate survivin expression and promote B-lymphoma cell proliferation (44). AlHV-1 encodes the largest genome maintenance protein described to date (24). It is, therefore, possible that AlHV-1 ORF73 might similarly have prooncogenic functions that could ultimately be involved in the triggering of CD8⁺ T-cell proliferation observed in MCF. It is unknown at this stage which ORF73 domains are involved in the genome maintenance function. However, the C-terminal domain shows similarities with the ones from other γ-herpesviruses, which is indicative of conserved genomic episome maintenance functions (18). The large internal acidic repeat domains that are conserved not at the sequence level but in regards to their structural properties are suggestive of conserved functions, such as cellular transcriptomic interactions and/or immune evasion properties that might be involved in MCF pathogenesis.

In this study, we provide strong evidence indicating that latency is at the heart of MCF. The mechanisms behind the onset of the pathology, however, are not fully understood but most likely, do not involve overabundance of lytic replication and consequent inflammation. Our bioluminescence analyses using 247N-luc⁺ WT and 73^null viruses suggested that ORF73 is not necessary for viral replication in vivo after primary infection and that AlHV-1 initially replicates in nasal turbinates and lungs. From then, we can hypothesize that, after entering the organism and infecting CD8⁺ T cells caused by AlHV-1 lymphotropism, the virus establishes latency in these cells. Although it is possible that the virus could only be a passenger in CD8⁺ T cells and trigger proliferation independently of latency-dependent mechanisms, our findings rather support that the deregulation and proliferation of CD8⁺ T cells have a direct link with latent infection. Indeed, we showed abundant AlHV-1 genomes that are essentially circular during MCF, identified high frequencies of infected CD8⁺ T cells, and showed that disruption of ORF73 rendered AlHV-1 apathogenic, whereas it did not significantly affect initial virus replication in vivo. According to this hypothesis, we can expect that latently infected CD8⁺ T cells present at the peak of the disease would result from the clonal expansion of only few mother cells infected after host colonization. Future experiments addressing this point need to be developed. For instance, it would be interesting to determine the T-cell receptor polymorphism of infected CD8⁺ T cells and whether only infected CD8⁺ T cells are proliferating in MCF. The exact mechanisms used by AlHV-1 to trigger proliferation remain to be deciphered. Although ORF73 could be directly involved in tumorogenesis, such as observed in KSHV malignancies (32), AlHV-1 also encodes numerous genes and predicted microRNAs of yet unknown functions that could be involved in processes involving, for example, cell cycle and/or apoptosis deregulation, mechanisms that are directly involved in malignancies induced by other γ-herpesvirus infection (32, 45–47). The identification of viral genes and regions expressed in MCF in our tiling array opens a perspective for future identification of such mechanisms. Additionally, microarray analyses of host gene differential expression identified numerous cellular genes that could be involved in the pathogenesis of MCF. Although the possible interpretations of these data are substantially limited, because it is not possible to distinguish a direct effect of AlHV-1 infection on gene expression from an indirect effect because of the disruption of the tissue itself (i.e., the difference in the cellular population phenotypes and proportions in Mock and MCF samples), we can conclude that MCF is associated with a strong disruption of the LN gene expression profile with up-regulation of proinflammatory cytokines, cytotoxic proteins, and genes involved in cell division. Additional investigations on sorted cell subsets, such as CD8⁺ T cells, should identify differential expression of genes directly caused by AlHV-1 infection and determine the involvement of such an inflammatory environment on the development of MCF lesions. Indeed, it has been shown in virus-induced malignancies, such as Kaposi sarcoma, that host cytokines and growth factors are necessary to ensure optimal tumorogenesis (32).

AlHV-1 infection of MCF-susceptible species is problematic in endemic regions and also zoos, where high value and endangered susceptible species are kept alongside wildebeest. Vaccination attempts have been performed with variable success (48–50). However, all strategies reside on the use of attenuated or inactivated WT virus, and full protection was unfortunately never obtained. We took advantage of the apathogenic 73^null virus to induce a protective response in the rabbit model. We obtained a strong antibody response and full protection after an intranasal challenge with a pathogenic WT virus. ORF73-deficient γ-herpesviruses have previously been shown to elicit effective protection against WT challenge (22, 34, 51) and have the unique advantage of being unable to persist. Although the specific protection mechanisms remain unknown, this recombinant AlHV-1 strain could induce a protective response in MCF-susceptible species and represent a serious candidate vaccine. Availability of such vaccine will bring prospects for effective protection against MCF for the local pastoralist population living in regions where cattle and wildebeest herds coexist.

Altogether, the findings detailed in this study suggest that AlHV-1 infection of MCF-susceptible species induces a peripheral T-cell lymphoma-like pathology, where ORF73 expression and latency are both essential.

Materials and Methods

Animals and Virus Infection in Vivo.

Eight healthy 4- to 6-mo-old Holstein–Friesan calves, free from Bovine herpesvirus 1 and 4 and Bovine Viral Diarrhea virus, were nurtured at the Veterinary and Agrochemical Research Center (Brussels, Belgium). Specific pathogen-free NZW rabbits were purchased from the Centre d’économie rurale (Marloie, Belgium) and used at 8–12 wk of age. Animals were inoculated i.v. with ∼3 × 10⁶ mock or AlHV-1–infected bovine turbinate (BT) cells. AlHV-1–infected cells were harvested when CPE reached ≥90% and normalized using ORF3 qPCR (SI Appendix, SI Materials and Methods). Animals were examined daily for clinical signs and rectal temperature. According to bioethical rules, calves and rabbits were euthanized after 48 h of persistent hyperthermia (>40 °C). The local ethics committees of the Veterinary and Agrochemical Research Center and the University of Liège for experiments involving calves and rabbits, respectively, have accredited the performed animal studies.

Microarray Design and Hybridization.

A custom microarray combining a high-density tiling array of the entire AlHV-1 genome with a set of probes targeting all bovine transcripts was designed for 8 × 60,000 glass slides (Agilent Technologies). Additional procedures are given in SI Appendix, SI Materials and Methods.

Viral Transcript Identification.

The delimitations of viral transcripts were inferred from the mean base pair log₂ fold change signal in the in vitro experiment (Fig. 1A). Because of the poly-dT amplification step, transcripts showed a characteristic shape defined by a sharp peak on their 3′ end. These peaks were identified using the wavelet transform as a multiscale shape detector in LastWave (http://www.cmap.polytechnique.fr/~bacry/LastWave). We used the first derivative of the Gaussian as the analyzing wavelet and determined the positions of the peaks as the abscissa to which the signal maxima converged. For each abscissa, we retained signal peaks (i) bearing a canonical polyadenylation signal within ±100 bp (AATAAA or ATTAAA) and (ii) with log₂ fold change that was above a 0.6 threshold. These peaks were used to define the 3′ ends of transcripts. Corresponding 5′ ends were defined manually from the remaining peaks and by checking the local signal shape. A few transcripts were manually added in regions that had failed to be detected by our wavelet-based method. The overall expression of transcripts was computed as the mean log₂ fold change along their length. Transcripts with overall expression that was above a 0.4 threshold were discussed in the text.

Production of AlHV-1 BAC Recombinant Plasmids and Viruses.

The AlHV-1 BAC clone was used to produce the recombinant plasmids using the galactokinase (galK) gene as selection marker in Escherichia coli (52). Additional procedures are given in SI Appendix, SI Materials and Methods.

Statistical Analysis.

Microarray analyses and figures were conducted using R and the limma package from the Bioconductor project (53–55) combined with ad hoc programs written in Python. All other statistical analyses were conducted using GraphPad Prism 4 software.

Additional procedures are given in SI Appendix, SI Materials and Methods.