The rabbit papillomavirus model: a valuable tool to study viral–host interactions

Abstract

Cottontail rabbit papillomavirus (CRPV) was the first DNA virus shown to be tumorigenic. The virus has since been renamed and is officially known as Sylvilagus floridanus papillomavirus 1 (SfPV1). Since its inception as a surrogate preclinical model for high-risk human papillomavirus (HPV) infections, the SfPV1/rabbit model has been widely used to study viral–host interactions and has played a pivotal role in the successful development of three prophylactic virus-like particle vaccines. In this review, we will focus on the use of the model to gain a better understanding of viral pathogenesis, gene function and host immune responses to viral infections. We will discuss the application of the model in HPV-associated vaccine testing, in therapeutic vaccine development (using our novel HLA-A2.1 transgenic rabbits) and in the development and validation of novel anti-viral and anti-tumour compounds. Our goal is to demonstrate the role the SfPV1/rabbit model has played, and continues to play, in helping to unravel the intricacies of papillomavirus infections and to develop tools to thwart the disease.

This article is part of the theme issue ‘Silent cancer agents: multi-disciplinary modelling of human DNA oncoviruses’.

1. The cottontail rabbit papillomavirus/rabbit model system: a historical perspective

Cottontail rabbits (Sylvilagus floridanus) of the midwest United States often manifest large warty structures on cutaneous skin sites [1] (figure 1). In 1933, Shope & Hurst [2] reported the isolation of an infectious agent from these structures. This agent retained its infectivity upon filtration and a virus responsible for the pathology was soon identified [3]. The virus was first called Shope papillomavirus (SPV) but later became known as the cottontail rabbit papillomavirus (CRPV) [4], and is now known as Sylvilagus floridanus papillomavirus 1 (SfPV1). Early studies focused on the progression of lesions to carcinomas and demonstrated that the virus could also infect domestic rabbits (Oryctolagus cuniculus) [5]. The latter discovery opened the door for a small animal model to investigate papillomavirus pathology. Among important findings in the early years were the following: (i) wild rabbit lesions were far more viral DNA and virus particle-rich than domestic rabbit lesions [6] (figure 1); (ii) naked viral DNA is infectious [7], including plasmid DNA cloned between the early and late regions of the viral genome [8]; (iii) tumours in domestic rabbits were more prone to cancers and a subset of these metastasized to the lung and other organs [7–10]; (iv) viral DNA copy number in benign lesions was correspondingly higher in wild rabbit lesions (figure 1) [9,10]. DNA in cancers, however, was hard to detect in both species, although viral transcripts could be found [11,12]; (v) sera from both SfPV1-infected wild and domestic rabbits could neutralize the virus in vitro and in vivo [6,13,14]; (vi) the genomic structure was determined. Strongest homologies were found with HPV1 [15,16]; (vii) viral transcripts have been mapped and the cap sites have been determined [17]. The virus encodes 10 genes, two of which (LE6, the corresponding protein is translated from the first open reading frame starting at 154 bp, with 272aa in length and E7) code for two essential transforming proteins [18,19], two of which are capsid proteins (L1 and L2) [14,20], two of which are involved in the control of replication and gene regulation (E1 and E2) [21–26] and two others, E9^E2C (renamed E8^E2) which has no effect on growth of the lesions but acts as a transcriptional repressor in mammalian cells [26] (https://pave.niaid.nih.gov/) and E4 which plays a role in DNA amplification and L1 expression [27] (figure 2); (viii) unlike LE6 and E7 that are essential for SfPV1 infection, two additional oncogenes SE6, the corresponding protein is translated from the first open reading frame starting at 445 bp, with 175aa in length and E8 (renamed E10) (https://pave.niaid.nih.gov/) both have an effect on tumour growth but not on viral production [31–33]; (ix) early genes E1 and E2 are also absolutely required for infection [24,28,34] (figure 2); (x) E6 and E7 are expressed in malignant lesions and have been posited to be targets for therapeutic interventions [17,31]; (xi) both episomal and integrated DNAs were found in SfPV1-induced carcinomas, and DNA was highly methylated in malignancies in domestic rabbits [32]; (xii) both progressive and regressive variants of the virus exist [33,35,36]; and (xiii) cells infected with SfPV1 were found to co-localize with hair follicle stem cells [37].

Figure 1.

Discovery of Sylvalagus floridanus papillomavirus 1 (SfPV1) and the development of the SfPV1/domestic rabbit model. Wild cottontail rabbits with long horny growths around the ears and chin and on the torso are common in the midwest (a). The growths are actually warts induced by papillomavirus infections. The virus isolated from these warts is infectious and induces tumour growth in the wild rabbit (b, arrows) and domestic New Zealand white (NZW) rabbits (c). Most tumours on domestic rabbits progress to cancer within a year or more of infection (c). Viral capsid protein and viral DNA are detected in infected tissues of cottontail rabbit lesions (d,e) and NZW domestic rabbit (f,g) by immunohistochemistry (IHC) and by in situ hybridization (ISH), respectively.

Figure 2.

The SfPV1 (CRPV) genome. Genes found to be essential to produce papillomas include LE6, E7, E1 and E2 [19]. Two late genes L1 and L2 are required for the viral life cycle [14]. Other genes including E4, SE6, E8 (renamed E10) [16,27] and E9 ^E2C (renamed E8 ^E2) [22] (https://pave.niaid.nih.gov/) are not essential for SfPV1 infection and tumour growth in vivo [28]. E5 is not considered a functional gene, although it showed transformation activity in vitro [29,30].

2. Investigation of virus–host interactions using the Sylvilagus floridanus papillomavirus 1/rabbit model

(a). Sylvilagus floridanus papillomavirus 1 pathogenesis

Much has been learned about viral pathogenesis using the SfPV1 model. Studies have been greatly facilitated by the fact that viral DNA is capable of producing infections [8,38,39]. Viral infections are initiated in the basal cells and thus must gain access to these cells via abrasions in the skin. Wounding has thus been recognized as essential for papillomavirus infections. Initially, our laboratory used a combination of turpentine and acetone to induce hyperplasticity in the skin and then worked the DNA (in the form of plasmid) or virus into the skin with a needle [8]. We subsequently investigated simple wounding without chemicals and developed a protocol that is far more benign for both technician and animal [40]. Using this protocol, we have achieved infectivity efficiencies orders of magnitude higher and results that are highly reproducible from experiment to experiment [41]. Other laboratories have employed a gene-gun to induce infections from DNA [35,42,43]. In these laboratories, the viral DNA—excised, re-ligated or in plasmid form—was coated onto the ‘bullets’ for delivery into the skin [27]. In our hands, this technique was less efficient and less reproducible; however, it has the advantage of delivering DNA that contains no extraneous intervening sequences [31].

Many studies have been conducted using plasmid DNA and much important information has been gleaned. A key feature of the system is that mutations can be easily inserted into the viral genome and functionally tested in vivo. Using this strategy, the E4 gene was found to be essential for the completion of the productive cycle, although not essential for infection [27]. The E5 gene start codon was mutated and found to be dispensable for infection [15,16]. The E8 gene was reported to be essential by one research group [44] and dispensable by another [16]; differences in methodology probably played a significant role in these outcomes [41]. SiRNA-expressing cassettes were engineered into the viral genome to monitor infectious outcome in situ [45].

Both progressive and regressive variants of SfPV1 are extant. Genetic mutations were used to determine that a 15 bp region at the C terminus of the E6 gene was responsible for this difference [36]. This region contains a PDZ binding domain that is important for the functioning of E6 [46]. Interestingly, a single mutation in wild-type (progressive) E6 that corresponds to one of the changes in the regressive E6 results in different outcomes in inbred and outbred laboratory rabbits [47]. Thus, genetic makeup of both host and virus play a crucial role in the outcome of infection.

Multiple SfPV1 mutants have been generated in our laboratory including many with mutations in the upstream regulation region [28]. All mutants have been tested in vivo to yield interesting insights into the function and plasticity of the SfPV1 genome and its interaction with the host. A novel ‘tandem repeat’ strategy has been established to enhance tumour growth in cases where mutations have created genomes that resulted in diminished tumour growth. The mutant genomes that were cloned with an additional fragment of SfPV1 demonstrated improved tumour growth when compared with those without this tandem repeat [28,48].

Papillomaviruses have developed multiple strategies to escape host immune surveillance. One is the use of rare codons [49–51]. To explore whether synonymous codon modifications in the SfPV1 genome would impact disease outcome in rabbits, we engineered a panel of mutants with synonymous codon changes in two oncogenes (E6 and E7) to make them more mammalian-like [52]. This strategy has been used to increase L1 expression of different human papillomavirus (HPV) types in vitro [53,54]. We detected dramatic phenotypic changes as a result of the codon optimizations [52]. Intriguingly, codon-modified E6 synergized with codon-modified E7 and induced early cancer development in domestic rabbits [52]. Some of these constructs, by contrast, showed higher regression rates than those of the wild-type [55]. These findings suggest that synonymous codon modifications can alter the immunogenicity of the virus and lead to either rapid progression or eventual regression depending upon the changes made. Our unpublished observations (N. M. Cladel 2008) include a cancer, which spontaneously regressed. Codon modifications of papillomavirus genes hold promise for future studies of host immune response, cancer development and viral gene functions. Synonymous mutations were once considered to be of little or no functional importance until recent studies demonstrated the significant function or phenotypical changes in difference model systems [56–59]. The findings reported here for SfPV1 contribute to the growing body of evidence that, in fact, synonymous changes can result in profound changes in gene expression.

(b). Host immune response to Sylvilagus floridanus papillomavirus 1 infections

Host immunity, including innate and adaptive immune responses, contributes to disease control as well as progression when circumvented [35,60–65]. A particular major histocompatibility complex class II (MHCII) genotype has been linked to the regressive phenotype for SfPV1 infections [66]. Further studies demonstrated that the rabbit MHCII (DRA–DQA) haplotype plays a role in tumour regression [35]. An inbred rabbit line (EIII/JC) mounts a vigorous response to SfPV1 infections and animals experience a regression rate of about 10%. By contrast, the outbred rabbits have a very low regression rate (about 1%) [47]. The human leukocyte antigens (HLA)-A2.1 transgenic rabbits also demonstrated higher immunogenicity when compared with outbred domestic rabbits [67]. Host T and B-cell-mediated immune responses are important factors in SfPV1-associated infections and diseases [60,64,68]. Tumour regression was correlated with infiltration of CD4 and CD8T cells that target early proteins E2 and E6 [60,69]. Taking advantage of these findings, prophylactic and therapeutic vaccines composed of DNA, peptides or protein were designed to target early genes as amplified below [31,70–72]. When host immune responses were suppressed with cyclosporine A, the regressive strain became persistent, supporting the observations that CD4 and CD8T cells play a critical role in controlling disease progression [47,69].

3. Prophylactic and therapeutic vaccine development using the Sylvilagus floridanus papillomavirus 1/rabbit model

(a). Prophylactic vaccines using the early and late genes as targets

A study in domestic rabbits using tumour suspensions to immunize animals demonstrated that tumour cells contained an antigen or antigens that could stimulate immune responses in the host [6]. Furthermore, sera harvested from these immunized animals neutralized the virions both in vitro and in vivo [73,74]. These studies demonstrated that the host could generate anti- SfPV1 antibodies and eventually led to the finding that immunization with L1 virus-like particles (VLPs) of rabbits could yield long-lasting protection from infection [75–77]. This finding, in turn, led to the development of the current L1 VLP vaccines, which are highly effective and which are projected to markedly reduce the incidence of HPV infection and subsequent cervical cancer in recipients [78]. In addition to stimulating strong humoral immune response, L1 protein was also found to elicit broad cellular immune responses in humans. The cellular immune responses resulting from the L1 vaccination have been further demonstrated in the rabbit model [79,80]. The L2 protein is far less immunogenic than the L1 protein, but has the advantage of producing antibodies that are cross-protective across a number of viral species [81,82]. Recent developments in pseudovirus and quasi-virus production have paved the way for the testing of prophylactic vaccines against multiple HPV types in the SfPV1/rabbit model [83,84]. We produced infectious hybrid viruses by encapsidating the SfPV1 genome into capsids of several oncogenic HPVs (HPV16, 18, 31, 45 and 58). All of these hybrid viruses were infectious in rabbits as determined by the generation of tumours [84,85].

Protective immunity by early gene products has been tested in the rabbit model as well as in other animal models for more than a decade. Owing to the poor immunogenicity of these proteins, low or undetectable antibody levels were found in most immunized animals [86,87]. By contrast, cell-mediated immune responses were found to play a critical role in the host protection against viral infections [63,86–89]. Interestingly, the method of delivering the antigen also plays an important role in the outcome [90–92]. Antigens can be delivered as protein products, peptides or as DNA and the outcomes can be different depending upon the mode of delivery. For example, DNA delivered by gene-gun promoted strong cell-mediated immune response while intramuscular injection of DNA offered no protection, even though some cell-mediated immune responses were generated in the immunized animals [87].

(b). Therapeutic vaccines using the early and late genes as targets

Although prophylactic vaccines provide excellent protection against new infections, they offer no protection against established infections. Therefore, a vaccine that could resolve active infections would be highly desirable. The SfPV1 rabbit model has been used to investigate effective targets for therapeutic purposes [17,72,93,94]. Because early genes promote cell-mediated immune responses in the host and hold promise to clear the infection, early proteins (E1, E2, E6 and E7) have been tested for their therapeutic potential in different formats. These include DNA vaccines delivered by gene-gun or by viral vector [31,95] (figure 3). E1 and E6 proteins have proved to be more effective for eliminating established tumours than E7 and E2, while the combination of the four early genes yielded the best results [70,89,96]. Recent promising studies used long peptides of E6 and E7 proteins as immunogens [72,94]. This immunization strategy showed a strong therapeutic effect in the SfPV1/rabbit model as well as in human clinical trials [94,97–99].

Figure 3.

Therapeutic vaccination with HLA-A2.1 restricted epitope DNA vaccines targeting early gene E1. Control (HLA-A2.1-negative) and HLA-A2.1 transgenic rabbits (HLA-A2.1-positive) rabbits are both positive for rabbit MHCI as shown by immunohistochemistry on the left. Animals infected with SfPV1 were divided into four groups and vaccinated with two different E1 epitope DNA vaccines (E1/303–311 and E1/161–169) at week 4 post-infection [95]. Two booster immunizations were conducted at three-week intervals (arrows). Significantly smaller tumours were found in the transgenic rabbit group immunized with epitope E1/303–311 when compared with the other three groups (by geometric mean diameter (GMD) in mm, p < 0.05, unpaired Student t-test) indicating the immune response was effective and specific.

4. The Sylvilagus floridanus papillomavirus 1/rabbit model for the testing of anti-viral compounds

The SfPV1/rabbit model is valued for its reproducibility and has been widely used for many years to test anti-tumour compounds [100,101]. Topical cidofovir treatment has been shown to be effective in resolving the tumours, but the recurrence rate is high [101] (figure 4). The effect can be improved by combining cidofovir treatment with DNA vaccination by gene-gun [102]. Our laboratory has tested many other compounds over the last several decades. Unfortunately, the efficacy was not optimal in most studies [55,103]. New strategies involving changes in methods of delivery, formulations and combination treatments hold promise in the coming years [101–105]. Our ultimate goal is to find ways to eradicate papillomavirus-associated diseases and cancers; it is our expectation that the SfPV1/rabbit model will continue to prove very useful in this quest.

Figure 4.

Therapeutic treatment with cidofovir and vehicle (cremophor) at weeks 2–4 post-viral infections [101]. Two groups of animals were infected with wild-type (wt.) SfPV1 (induces large tumours) and SfPV1E8ko (induce small and persistent tumours) at two back sites, respectively. Two tumours induced by either wt. SfPV1or SfPV1E8ko on the left (L) side of the rabbits were treated with either 1.0% cidofovir formulated in cremophor (a) or cremophor (b) from week 2 to week 4 post-infection. The right (R) two sites were not treated and served as controls for the treated sites on the same animal. The animals were monitored for tumour growth weekly and pictures were recorded. Cidofovir-treated sites were free of tumours induced by both wt. SfPV1 and SfPV1E8ko at week 7 post-infection, while no reduction in tumours was found on the vehicle-treated sites (p < 0.05, unpaired Student t-test).

5. Limitations and future directions

The SfPV1/rabbit papillomavirus model induces infection with long-term persistence and malignant progression of lesions both consistently and reproducibly. Eighty years of studies with this unique model have proved it to be one of the best surrogate models for studying high-risk HPV infection [55,100]. Over the past several decades, we and others have generated a large panel of mutants to aid in the study of the pathogenesis of papillomavirus [55]. These mutant constructs will allow us to dissect the function of different genes and their role in the pathogenesis of papillomavirus-associated skin infections. However, SfPV1 only infects cutaneous sites and therefore is not the ideal model to mimic anogenital infections and diseases. For future directions, we look forward to more comparative studies with human samples. These additional studies will help to uncover mechanisms that could ultimately lead to a novel treatment for HPV-associated diseases and cancers.

Data accessibility

This article has no additional data.

Competing interests

We declare we have no competing interests.

Funding

Research reported in this publication was supported by the National Cancer Institute under award number R01 CA47622 (N.C.), NIH contract HHSN272201000020I (N.C.) and the Jake Gittlen Memorial Golf Tournament.