Systemic Polyomavirus Genome Increase and Dissemination of Capsid-Defective Genomes in Mammary Gland Tumor-Bearing Mice

Abstract

BALB/c mice that developed tumors 7 to 8 months following neonatal infection by polyomavirus (PYV) wild-type strain A2 were characterized with respect to the abundance and integrity of the viral genome in the tumors and in 12 nontumorous organs. These patterns were compared to those found in tumor-free mice infected in parallel. Six mice were analyzed in detail including four sibling females with mammary gland tumors. In four of five mammary gland tumors, the viral genome had undergone a unique deletion and/or rearrangement. Three tumor-resident genomes with an apparently intact large T coding region were present in abundant levels in an unintegrated state. Two of these had undergone deletions and rearrangements involving the capsid genes and therefore lacked the capacity to produce live virus. In the comparative organ survey, the tumors harboring replication-competent genomes contained by far the highest levels of genomes of any tissue. However, the levels of PYV genomes in other organs were elevated by up to 1 to 2 orders of magnitude compared to those detected in the same organs of tumor-free mice. The genomes found in the nontumorous organs had the same rearrangements as the genomes residing in the tumors. The original wild-type genome was detected at low levels in a few organs, particularly in the kidneys. The data indicate that a systemic increase in the level of viral genomes occurred in conjunction with the induction of tumors by PYV. The results suggest two novel hypotheses: (i) that genomes may spread from the tumors to the usual PYV target tissues and (ii) that this dissemination may take place in the absence of capsids, providing an important path for a virus to escape from the immune response. This situation may offer a useful model for the spread of HPV accompanying HPV-induced oncogenesis.

The pattern of infection of murine polyomavirus (PYV) in its natural host has been extensively studied. When mice are infected at the neonatal stage, replication takes place at high rates in many tissues and systemic infection is rapidly reached (7, 9, 22, 25, 30). Following the development of an antiviral immune response at around 2 weeks of age, virus and infected cells are cleared from many tissues, causing a dramatic decrease in the state of infection, as measured, for example, by in situ hybridization of whole-mouse sections (7, 9, 22, 25, 26, 30). From this stage onward, a state of persistent infection is maintained, in particular in a restricted set of organs, namely, the kidneys, skin, bone, and mammary glands (9, 22, 30, 31). With time, the load of viral genomes in persistently infected tissues continues to decrease slowly (31) and the virus titer in mouse tissues decreases by 6 log units between 30 and 60 days (25). This is due in part to the continued process of immune clearance (25) and to an age-linked loss of replication potential in all tissues (6, 30, 31). This general pattern of infection, clearance, and persistence is modulated by many factors, such as the virus strain, the timing of infection, the route of infection, the host genetic background, and, modestly, the virus dose (1, 9, 10, 19, 22, 24, 26, 30). Under the conditions of the present study, BALB/c mice were infected peritoneally or subcutaneously with approximately 5 × 10⁶ PFU of wild-type A2 virus within 24 h of birth. In this case, the level of genomes in the persistence target tissues reached an overall level of less than one genome per cell by 20 weeks postinfection (31).

In time, neonatally infected mice may develop tumors. However, the tumor incidence is specific for each mouse strain. Owing to a strong immune response, most strains are resistant to PYV-induced oncogenesis. In the resistant strains, tumors are rare and appear only after a long delay (3, 20, 21). The BALB/c strain used in the present study shares these characteristics (29). The factor(s) responsible for the development of tumors in the presence of an immune response has not been determined so far. The frequent targets of oncogenesis are the organs in which the viral genome retains some replication potential at the adult stage (i.e., mammary gland, skin, and bone) (31). This suggests that continued replication plays a role in oncogenesis and that tumors may arise from delayed transformation events accompanied by some escape from immune surveillance.

In the present study, neonatally infected BALB/c mice were monitored for the development of tumors. Six tumorous mice were studied in detail with respect to the load of viral genomes in the tumors and in 12 organs. The levels of genomes in the nontumorous organs of these mice were found to be elevated compared to those found in the same organs of tumor-free mice infected in parallel for the same length of time. Because the viral genomes in four tumors had acquired unique deletions, it was possible to track the tumor-resident genomes. The results suggest that the systemic elevation in the levels of genomes was caused at least in part by the spread of viral genomes that originated in the tumors. In two cases, the rearrangements resulted in an inability to encode capsids and hence to produce live virus. Therefore, we entertain the possibility that the systemic spread can be achieved by the movement of nonencapsidated viral genomes.

MATERIALS AND METHODS

Virus, mice, and infections.

Stocks of PYV wild-type strain A2 (WTA2) (12) were grown either in mouse NIH 3T3 cells or in primary baby mouse kidney cells. Midpregnancy BALB/c mice were purchased from Harlan, Sprague-Dawley. Within 24 h of birth, neonatal mice were infected intraperitoneally with 5 × 10⁶ PFU of WTA2. The mice were monitored weekly for the development of tumors. Mice were sacrificed when they developed pathological symptoms, when the tumors reached 1.5 cm in diameter (mammary gland tumors), or at the times shown. All remaining mice were sacrificed at 60 weeks. All the mice were euthanized by treatment with CO₂.

Analysis of viral DNA.

Briefly, as described previously (30), mice were sacrificed when tumors developed or at the times described in the figure legends or in the text. Tumors or organs were removed, homogenized, and digested overnight in a buffer solution containing proteinase K. Total DNA was extracted and isolated using standard techniques. DNA samples (5 to 20 μg) were digested with various restriction endonucleases. Agarose gel electrophoresis was carried out with 0.7 to 1% agarose gels when using enzymes that cut the genome once (EcoRI) or not at all (BglII) or with 1.5 to 2% agarose gels following treatment with nucleases that cut the genome in multiple locations (e.g., MspI). DNA was transferred to Hybond filters and hybridized to ³²P-labeled probes for 72 h at 65°C. ³²P-labeled probes were synthesized using the Amersham multiprime labeling kit and protocol, using linearized WTA2 genome or cellular murine immunoglobulin μj gene DNA as the template. The specific activity of the probes was always higher than 10⁹ cpm per μg of DNA. After being washed, the blots were exposed to Hyperfilm MP (Amersham). Counts in the bands were determined with a phosphorimager (AMBIS) either as direct counts or as film band intensities. To compensate for uneven loading of wells, the data were computed as the ratio of the PYV signal to that of the cellular μj signal. To estimate the average genome copy number per cell, DNA isolated from a cell line that contains one integrated copy of the PYV genome was included in some of the blots.

Analysis of viral RNA.

Briefly, as described previously (14), total RNA was isolated from organ homogenates in a solution containing guanidinium isothiocyanate. Samples (20 μg) were electrophoresed in formaldehyde-containing gels and blotted onto Nytran membranes. The blot was hybridized first with ³²P-labeled PYV genomic DNA, which detects all mRNA species. To control for evenness in loading, the blots were stripped and rehybridized with a ³²P-labeled probe for the cellular glyceraldehyde 3-phosphate dehydrogenase (GAPDH) mRNA.

PYV late mRNA species were detected with late specific probes. To score for VP1-specific mRNA, PYV sequences spanning nucleotides (nt) 3918 to 2928 (Fig. 1) were inserted between the HindIII and BamHI sites of pSPT18 (Roche Molecular Biochemicals). The plasmid was cleaved with HindIII, and the T7 RNA polymerase was used to synthesize the anti-VP1 probe. To detect the VP2/VP3 mRNA, we used a pGEM-1-based plasmid that contains PYV MspI fragment 3 (Fig. 1) cloned between the HindIII and EcoRI sites. The plasmid was cleaved with EcoRI, and the SP6 RNA polymerase was used to synthesize the anti-VP2/VP3 probe. The anti-late-mRNA probes were labeled with digoxigenin (DIG) with the kit from Roche Molecular Biochemicals. Hybridizations with DIG-substituted probes were carried out as follows. Stripped blots were prehybridized for 2 h at 68°C in 50% formamide–5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)–0.02% sodium dodecyl sulfate (SDS)–0.1% N-lauroylsarcosine–2% blocking reagent (Roche Molecular Biochemicals) and then hybridized overnight at 68°C in a hybridization mixture containing DIG-labeled RNA probe. The membranes were washed twice with 2× SSC–0.01% SDS at room temperature for 15 min each and twice with 0.5× SSC–0.01% SDS at 68°C for 15 min each. The membranes were treated with blocking agent solution for 1 h and then with anti-DIG-AP, diluted 10,000-fold in blocking buffer, for 30 min. After extensive washing, the chemiluminescent phosphatase detection substrate CSPD was applied for 1 min and the membranes were exposed to X-ray films.

FIG. 1.

Map of the PYV genome. A physical map of the genome is shown. Various landmarks are included: an inner circle with the position of useful restriction endonuclease sites (MspI and its eight fragments and the unique EcoRI, BglII, and BclI sites); nucleotide position markers every 500 nt; the origin; the enhancer region, which controls the expression of both early and late transcripts; the three early mRNAs and their proteins large, middle and small T antigens; and the three late mRNAs and their three proteins VP1, VP2, and VP3, with the position of the 5′ and 3′ splice sites for the VP1 transcript.

Virus assay.

A portion of the organ or tumor homogenate was used to assay for the virus as follow (31). Homogenates were sonicated for 1.5 min at full power with a sonic oscillator (250 W, 115 V, 10 KC, 60 cycles). The lysates were then incubated for 15 min at 45°C and centrifuged at room temperature at 2,000 rpm for 20 min in a Sorval GSA rotor. Supernatants were removed and stored at −20°C until use. Titers were determined by plaque assay on NIH 3T3 cells.

RESULTS

Pattern of tumor development in neonatally infected BALB/c mice.

BALB/c mice were infected intraperitoneally within 24 h of birth with PYV WTA2 and maintained as described in Materials and Methods. Mice remained caged with littermates of the same sex for the duration of the experiment. Half of the mice were sacrificed during the first 20 weeks postinfection to determine the patterns of viral genome spread and persistence. No tumors had developed by that time. The results were published previously (31). The 38 remaining mice, representing five litters, were examined weekly for approximately a year for the induction of palpable tumors, and the results for these mice are presented here.

During weeks 28 to 31 postinfection, mammary gland tumors became evident in four of six females from one litter (L1). These mice are designated F1¹, F2¹, F4¹, and F5¹. One of these mice, F4¹, developed two independent tumors in the L2 and R2 mammary glands. A fifth female from the same litter (F3¹) developed a hemangioma in the same period, while the sixth (F6¹) was tumor free (Table 1). All six females from this “tumor-prone” L1 litter were sacrificed within the same period (between weeks 28 and 31). The two males from the same litter (M1¹ and M2¹) were kept alive and remained tumor free until the termination of the experiment at 60 weeks.

TABLE 1.

Distribution of tumors^a

Mouse type	Litterb	No. of micec	Time (wk)d	Tumor type
Tumorous mice
Females	L1	4	28–31	Mammary gland tumor
		1	28	Hemangioma
Males	L2	1	30	Bone tumor
	L3	1	37	Bone tumor
	L4	1	57	Bone tumor
Tumor-free mice
Females	4	11	60	None
Males	5	15	60	None

^aThirty-eight neonates born to six BALB/c mice were infected intraperitoneally within 24 hours of birth with 5 × 10⁶ PYV WTA2 and examined weekly until 60 weeks for the induction of palpable tumors. In the L1 litter of eight pups, mammary gland tumors and a hemangioma developed in four females and one female, respectively. Three males from three independent litters (designated L2, L3, and L4) each developed an osteosarcoma. 

^bThe specific litter in which tumor induction was observed is indicated for the tumorous mice. The number of litters with tumor-free mice is shown for the tumor-free mice. 

^cThe number of mice with tumors is indicated for the tumorous mice; the total number used is indicated for the tumor-free mice. 

^dThe time of appearance of the tumors is shown for the tumorous mice. Mice were sacrificed within 2 to 3 weeks after the appearance of the tumors. The time of sacrifice is shown for the tumor-free mice. Healthy mice were observed for over a year and sacrificed at the end of the experiment, except for one female (F6¹) from the L1 litter and two males (M4 and M5), which were euthanized at the same time as its sibs (F6¹), or at 34 weeks (M4 and M5). 

In the other four litters, 3 of 21 males from three different litters developed bone tumors at 30, 37, and 57 weeks of age. In two cases, these mice developed hindleg paralysis, presumably as a consequence of spinal cord compression by the vertebral location of the tumor (13). One male developed an enlarged bladder and was sacrificed at 30 weeks of age. Two healthy males were sacrificed at 34 weeks for additional time points in the analysis of genome persistence. Eleven females originating from four litters and 15 males from five litters remained tumor and disease free to the end of the experiment.

Analysis of the viral genomes in the mammary gland tumors.

A restriction endonuclease analysis of the viral genomes in the mammary gland tumors was carried out. Figure 2 shows the results of a digest with BglII, an enzyme that does not digest the PYV genome and therefore reveals unintegrated and integrated genomes. Tumors T1, T2, and T5 contained very high (T1 and T2) or high (T5) levels of unintegrated genomes. In tumors T4a and T4b, the PYV genome was present in an integrated state at only a single site; as shown below, in both cases the genome was defective and thus the genome copy number was <1 copy/cell. In comparative blots including standards with known genome copy numbers, the levels of viral genomes in the tumors were estimated to average approximately 50 copies per cell for T1 and T2 and 10 copies/cell for T5 (data not shown). As supported by results presented in Fig. 3, 5, and 6, for T1 and T2 the two major bands represent the supercoiled (bottom band) and relaxed circular forms (top band) of a single molecular species. In T5, two species of different sizes were present. The genomes in tumors T1 and T5 were shorter than those in tumor T2. Other larger bands representing either multimeric or integrated viral sequences were also seen. For mouse F3¹, the DNA from the hemangioma could not be recovered.

FIG. 2.

Analysis of the unintegrated viral genomes in mammary gland tumors. Tumors T1, T2, T4a, T4b, and T5 were collected and total DNA was extracted and processed as described in Materials and Methods. DNA was digested with BglII, a restriction endonuclease that does not cut the PYV genome. Electrophoresis, blotting, and hybridization conditions are described in Materials and Methods. Lanes are designated by the tumor numbers. In lanes 1 and 2, the top band is the relaxed circular DNA form and the bottom band is supercoiled DNA.

FIG. 3.

Analysis of the intactness of the viral genomes in mammary gland tumors. DNA from the tumors was processed as described in the legend to Fig. 2. DNA was digested with MspI, an enzyme that cleaves the PYV genome into eight fragments. Lanes are designated by the tumor number. (A) Blot hybridization with a PYV genomic probe. Different film or scan exposures were used. The exposure for lanes 4a and 4b was five times as long as that for lanes 1, 2, and 5. Scan exposures for lanes 1 and 2 were shorter than those for lanes 5, 4a, and 4b. The positions of wild-type MspI fragments were determined by the use of marker DNA and are shown as numbers 1 to 7 on the right. (B) Identification of viral sequences in the MspI-generated fragments. The blotted membrane from panel A was serially rehybridized to probes specific for wild-type MspI fragments 1 to 4. The qualitative results for these four hybridizations are shown in an integrated schematic manner. The fragment-specific sequences are represented as follows: Fg1, ⊖; Fg2, ; Fg3, ①; Fg4, . Fragments with dual symbols represent examples of hybridization of two fragment-specific probes to the same fragment. Fragments represented at submolar levels are shown in brackets.

FIG. 5.

Test for the presence of viral genomes in nontumorous organs of tumor-bearing mice. Total DNA from 12 organs was extracted and processed as described in the legend to Fig. 2. The DNA was treated with EcoRI, which has a single site in the PYV genome and linearizes it. Seven mice were analyzed as shown on the right: the six females from the L1 tumor-prone litter, including four mice with one (F1¹, F2¹, and F5¹) or two (F4¹) mammary gland tumors (the two tumors in F4¹ are designated 4a and 4b), one mouse with a hemangioma (F3¹), and one tumor-free mouse (F6¹); as well as one male from a different litter with an osteosarcoma (M3²). p, pancreas; r, bone (rib); sp, spleen; lv, liver; lu, lungs; k, kidneys; sk, skin; b, brain; sg, salivary gland; mg, nontumorous mammary glands; ov/ts, ovaries (testes in mouse M3²); h, heart; t, tumor.

FIG. 6.

Qualitative comparisons of persistent and tumor-resident genomes. DNA from 12 organs of tumor-bearing mice was analyzed as described in the legend to Fig. 5. Analysis was carried out with MspI. Organs are denoted as in the legend to Fig. 5. The positions of wild-type fragments are shown on the left. (A) Analysis of organs from mouse F1¹. (B) Analysis of organs and the two tumors, 4a and 4b, from mouse F4¹.

Further analysis of the integrity and structure of the viral genomes was carried out with MspI, an enzyme that cleaves the PYV genome into eight fragments referred to as Fg1 to Fg8 (Fig. 1). The results are shown in Fig. 3A. The expectation was that the coding information for the middle T oncogenic protein would be present in all tumors. This protein encompasses MspI fragments Fg5, Fg4, and Fg7. These fragments were present in all five tumors. The enhancer sequences, which control the expression of the early and late proteins, were also expected. These map in MspI Fg3, specifically in the origin-proximal 200 bp (nt 5110 to 5273). A normal-size Fg3 was present in tumor T2 and apparently in T5, present at submolar levels in T4a, and missing in T1 and T4b. The enhancer-spanning fragment produced by codigestion with BglI and BclI (nt 89 to 5023) was present in T4a, T4b, in addition to T2, but not in T1 and T5 tumors (data not shown). Further analyses of the Fg3 sequences are presented below. An intact DNA replication function was expected to be present in tumors T1, T2, and T5, since these tumors contained high levels of unintegrated viral genomes. MspI Fg2, which encodes a major domain of large T antigen near the carboxy terminus, was indeed present in these tumors. Fg2 was missing altogether in T4b and was present at a submolar level in T4a. Finally, the capsid encoding Fg1 was either missing, present at submolar levels, or obviously altered in tumors T4b, T4a, and T5, respectively.

To further examine the identity of the various MspI genome fragments, the blot was rehybridized sequentially with probes specific for the four largest fragments, Fg1 to Fg4. The results are summarized in Fig. 3B. From these and other analyses, the following conclusions can be drawn for the five tumor-resident genomes.

In tumor T2, the fragment-specific hybridization analysis confirmed that the PYV genome contains fragments Fg1 to Fg4 of normal sizes. This tumor appears to harbor a wild-type genome. This was the only case among the five mammary gland tumors.

In tumor T1, hybridization with a genomic probe demonstrated a complete absence of the 882-bp long Fg3 fragment. Rehybridization with an Fg3-specific probe revealed the presence of Fg3 sequences in the fragment migrating slightly slower than Fg1. This aberrant fragment also contained Fg1-specific sequences. Given the size of the T1 genome deduced from BglII and EcoRI analyses (Fig. 2; see Fig. 5, below), we conclude that the T1 genome contained a deletion of slightly less than 882 bp, which spans the junction between Fg1 and Fg3. The deletion eliminated the MspI site at nt 4411 and fused contiguous Fg3 and Fg1 sequences. Since the BclI site (nt 5023) was also lost, the deletion is likely to span approximately nt 5023 and 4123. The VP1 5′ mRNA splice site at nt 5018 to 5022 and the 3′ site at nt 4122 to 4125 may be affected or removed. The deletion left the VP1 coding region intact (nt 4076 to 2928) but eliminated the VP2 and VP3 coding sequences. These proteins play a crucial role in viral dissemination (27).

In tumor T5, Fg3-specific sequences were detected in a fragment migrating very similarly to Fg3; this fragment also contained Fg1-specific sequences. Additional Fg1-specific sequences were detected in the other aberrant fragment migrating between Fg1 and Fg2. Thus, in the T5 genome, a rearrangement involving Fg3 and Fg1 sequences occurred, resulting in a net loss of approximately 200 bp. Other data (e.g., the presence of two BclI sites [data not shown]) suggest that rather than a simple deletion, a rearrangement took place involving a reiteration(s) and could therefore be accompanied by a deletion larger than 200 bp. Similarly to the T1 genome, the T5 genome has the coding capacity to replicate its genome and to transform cells but not the capacity required for the formation of normal viral particles.

In tumors T4a and T4b, Fg3-specific enhancer sequences were found in a fragment migrating with mobility very similar to that of Fg2. The PYV genomes integrated in these two tumors were very defective, encompassing essentially the minimal transforming region: the middle/small-T coding region (Fg5, Fg4, and Fg7) and the enhancer (BclI-BglI fragment). These genomes, which lack Fg2, are incapable of replication. Their maintenance was accomplished by integration into the host genome. The apparent similarity of the viral sequences in the T4a and T4b genomes may reflect an accidental fragmentary integration in both tumors or the integration of a circulating defective variant that caused these two independent tumors. In addition, tumor T4a contained normal-size Fg1, Fg2, and Fg3 fragments, present at submolar levels. As discussed below, this is likely to reflect the presence of a wild-type genome in tumor T4a. Since the abundance of the wild-type sequences was lower than that of the integrated variant, the wild-type genome could not be present in all tumor cells and may have resided in stromal or other components of the tumor.

Analysis of viral transcripts in the tumors.

Total RNA was isolated from the mammary gland tumors, and viral transcripts were examined by Northern blotting using the whole PYV genome as a probe. The analysis revealed quantitative and qualitative differences in the transcripts expressed in different tumors (Fig. 4A). The levels of transcripts were roughly proportional to the levels of genomes: high levels in tumors T1, T2, and T5 and low level in tumors T4a and T4b. Rehybridization with a probe for the cellular GAPDH transcripts showed evenness in sample loading (Fig. 4D). The transcript pattern was complex in tumors T2 and T5 and simple in T1. The bulk of the transcripts migrated between the 18S and 28S rRNAs.

FIG. 4.

Analysis of viral transcripts in the tumors. Total RNA was isolated from sections of the tumors and processed for Northern blotting as described in Materials and Methods. (A) Hybridization with a ³²P-labeled PYV genomic probe. (B) Hybridization with a DIG-labeled anti-Fg1-specific probe. (C) Hybridization with a DIG-labeled anti-Fg3 specific probe. (D) Hybridization with a ³²P-labeled GAPDH-specific probe. The positions of the 18S and 28S rRNAs are shown. The unique and major Fg1-specific late band present in tumors containing a wild-type genome is designated L. The aberrant corresponding band in T1 is designated L′.

Two anti-late probes were used to specifically identify the VP1 and VP2/3 transcripts (see Materials and Methods). The late PYV mRNAs are of heterogeneous sizes because of transcription around the genome, accompanied by leader-to-leader splicing. The sizes of the basic mRNAs with a single leader are 16S for the VP1 mRNA and 19S and 18S for the VP2 and VP3 mRNAs, respectively. However, mRNAs with a single leader are reported to not be spliced efficiently and to be degraded (17).

The RNA patterns in tumors T2 and T5 were very similar. The Fg1/VP1-specific probe detected a major and a minor transcript, designated L and S (Fig. 4B). The L band was considerably larger than 18S, while the S band was slightly larger than the 1.2-kb GAPDH band, a size consistent with the 16S mature VP1 mRNA. In T2, the L transcript was abundant. This transcript was also observed in T5, albeit at lower levels, and in T4a, at yet lower levels. The minor S transcript was detected only in T2. The Fg3/VP2/VP3-specific probe detected the same L and S bands (Fig. 4C). In addition, in T2 and at lower levels in T5, other Fg3-containing transcripts were observed. These are likely to represent transcript species with a variable number of leaders in which Fg3 sequences are overrepresented (17).

A single species of late transcript was found in T1 that did not correspond to the size of any late transcripts present in T2. This transcript, L′, was shorter than the L species. It hybridized strongly with the Fg1/VP1 probe and weakly with Fg3 probe. Thus, mRNA containing the Fg3 VP2/VP3 sequences was of aberrant size and vastly underrepresented in this tumor.

Viral genomes in nontumorous organs of the tumor-bearing mice.

As described in the introduction, one hypothesis for the occurrence of breakthrough tumors in resistant mice is that it is caused by a decrease in the antitumor response. In turn, this might be reflected by a systemic increase in the distribution of virus and infected cells. Therefore, the presence of viral genomes in the tumor-free organs of the five tumor-bearing female mice (F1¹ to F5¹) and one tumor-free female mouse (F6¹) from the same tumor-prone litter was examined. Twelve organs were chosen: pancreas, bone (ribs), spleen, liver, lungs, kidneys, skin, brain, salivary glands, tumor-free mammary glands, ovaries, and heart. The results of an analysis with EcoRI, an enzyme with a single site in the PYV genome, are shown in Fig. 5 (and summarized in Table 2). Viral genomes were detected in many organs of the tumor-bearing mice. The mice with the highest levels of genomes in the tumors displayed correspondingly higher levels of genomes in their tumor-free organs. This point is examined more quantitatively below (see Fig. 7). For example, viral genomes were detected in all organs studied of mouse F1¹, in most organs of mouse F2¹, and in few organs of mice F5¹, F4¹, F3¹, and F6¹. Although the levels of genomes varied from organ to organ in each mouse, a pattern could be identified. The highest levels were observed in the bone, for which all mice tested were positive. High levels were also observed in the pancreas, while intermediate levels were found in the heart, kidneys, skin, and ovaries and low levels were found in the spleen, liver, lungs, brain, salivary glands, and tumor-free mammary glands. In the F1¹, F2¹, and F5¹ mice, the tumors represented the tissue with the highest levels of genomes. In contrast, in mouse F4¹, the levels of genomes were higher in the nontumorous organs, suggesting that their abundance was higher than in the T4a and T4b tumors (i.e., >0.4 genome/cell). This could not be determined in mouse F3¹, since the hemangioma was not recovered. The same analysis was carried out with a male mouse, M3², from a different litter that had developed an osteosarcoma and was sacrificed at 28 weeks. Similarly to the situation with the tumor-bearing mice from the tumor-prone litter, viral genomes were recovered in all 12 tissues examined (Fig. 5, M3²). Abundant genomes of wild-type size were detected in the bone tumor (data not shown).

TABLE 2.

Viral genomes in tissues of mice^a

Mouse	PYVc	Level of genomes inb:
		Tumor	Rib	Kidney	Skin	MGd	SGd	Pancreas	Spleen	Liver	Lung	Brain	Ovary	Heart
F11	Def	+++++	++++	±	++	+	+	+++	++	++	+	++	+++	++
	WT	−	−	+++	+	−	−	−	−	−	−	−	−	++
F21	WT	+++++	++++	++	++	+	±	++	+	±	+	±	+	+
F51	Def	+++	+++	+	++	±	+	+++	−	±	±	±	±	±
	WT		±	+
F41	Def	+(b)/+(a)	−	−	−	−	−	−	−	−	−		−	−
	WT	−(b)/+(a)	++	++	+	−	−	−	−	−	−		±	+
F31	WT	NA	+	++	±	±	±			±	±	±	−	+
F61	WT	NA	±	+	+	±	−		−	−	−	−	±	−

^aThe presence of viral genomes was determined by Southern blotting of total DNA extracted from the organs listed, as described in the legends of Fig. 3 to 5. The data from Fig. 3 to 5 are summarized, as well as those from other exposures of the same blots or from other blots not shown. 

^bThe intensity of the bands, as gauged by visual examination, was given a value, reflected by the number of plus signs. This number reflects differences in intensities, which are approximate and not intended to be linear. A minus sign indicates that the level was below the detection level. These intensity values are consistent within a given blot but not between different blots. 

^cWhere applicable (F1¹, F4¹, and F5¹), the genotype (wild type [wt] or tumor-specific defective [Def.]) is indicated. In mouse F4¹, the two independently induced tumors are indicated as a and b, and in these cases, the tumor-resident genomes were integrated. 

^dMG, mammary gland; SG, salivary gland. 

FIG. 7.

Quantitative comparison of genome levels in tumorous and tumor-free mice. DNA from mouse organs was treated with EcoRI and processed as described in the legend to Fig. 5. The mice analyzed were tumor-bearing mice F1¹, F2¹, F3¹, F4¹, and F5¹, a pool of four tumor-free mice sacrificed at 20 weeks (denoted P20), and one male, M4, sacrificed at 34 weeks. Organs are denoted as in the legend to Fig. 5. Hybridizations with a probe for the cellular μj gene and for the PYV genome were carried out sequentially. The positions of the bands representing wild type (WT) and the T1-specific (Def) genomes are marked with arrows.

The EcoRI analysis showed clearly that the sizes of the viral genomes present in all nontumorous organs studied in mouse F1¹ were identical to that of the defective T1-resident genome except in the kidneys. In the kidneys, a preponderant band of wild-type size and a very faint band of T1-specific size could be detected. A low level of wild-type-size genomes was also detected in the heart, skin, and bone, in addition to a preponderant T1-specific-size genome in these organs. Similarly, in mouse F5¹, the prevalent species in most organs was also apparently identical to that of the defective tumor variant (see below). In contrast, the tumor-free organs of mouse F4¹ contained only wild-type-size genomes.

To better examine the structure of the genomes in the tumor-free organs and their similarities to the tumor-resident genomes, an analysis was carried out with MspI, and the results are shown in Fig. 6. As in the experiment in Fig. 3, the defective viral genome in tumor T1 in mouse F1¹ lacked fragment Fg3 and displayed a fragment slightly larger than Fg1. An identical fragment pattern was found in all organs examined except the kidneys, heart, and skin (Fig. 6A). In the kidneys, only genomes with a wild-type-like pattern could be detected. The T1-specific variant, which was present as a very-low-abundance band in the EcoRI analysis, could not be detected in the MspI analysis. This is probably explained by the size difference of the diagnostic fragment produced by these two enzymes, i.e., 5.3 versus 0.88 kb. The skin and heart displayed a mix of wild-type and tumor-resident variant. In mouse F5¹, genomes with the signature pattern of the tumor-resident variant were found in the pancreas and the ribs. A low level of wild type was also detected in the ribs. In the kidneys, the predominant species was wild type with a low level of variant. The wild type was also detected at low level in the heart (data not shown). In mouse F4¹, genomes with a wild-type pattern were found in the ribs, lungs, kidneys, and skin (Fig. 6B). A rehybridization of the blot with an Fg3-specific probe showed that the altered fragment of Fg2-like size that contains Fg3 sequences was present only in the T4a and T4b tumors (data not shown).

The 12 nontumorous organs of mouse F1¹ and its T1¹ mammary gland tumor, which contained very high levels of defective viral genomes, were tested for the presence of live virus. No live virus was recovered from the tumor or from any of the organs.

Relative viral genome persistence in tumor-bearing and tumor-free mice.

In the previously published study of the 20-week time course persistence of the viral genome in this group of mice (31), we showed that the level of genomes diminished progressively with time. By 20 weeks postinfection, the level of persistent genomes was at the limit of detection by Southern blotting. The results presented above (Fig. 5 and 6) on the persistent genomes in organs of tumorous mice sacrificed between 29 and 32.5 weeks postinfection suggest that the level of persistent genomes was higher in the tumor-bearing mice than in the tumor-free mice. To examine this point, genome levels in tissues of tumor-free mice sacrificed at 20 weeks (a pool of four mice) or at 34 weeks (two individual mice) were directly compared to those found in the female mice of the tumor-prone litter. A Southern blot analysis of EcoRI digests of DNA from various organs (kidneys, tumor-free mammary glands, rib, skin, and salivary glands) was carried out. The blot was sequentially hybridized with a PYV genomic probe and one for the μj cellular gene (Fig. 7). As previously published, for nontumorous mice sacrificed at 20 weeks postinfection, faint bands could be detected in the bone and salivary glands. A faint band was also detected in the overloaded kidney and salivary-gland sample of one of the healthy males (M5) sacrificed at 34 weeks postinfection (data not shown). The optical band density of the PYV and μj signals was determined, and the relative PYV/μj band density is presented in Table 3. For these tumor-free animals, the ratio of μj to PYV hybridized signal was high, indicating that the PYV genome is present at less than one copy per cell. Similarly, the organs of the F6¹ tumor-free female from the tumor-prone litter showed a μj-to-PYV signal ratio of >1. A different pattern was observed in the four tumor-bearing females, which were sacrificed at 31 weeks postinfection. The ratio was clearly reversed in one or more organs of these tumor-bearing females with either mammary gland tumors (F1¹, F2¹, F4¹, and F5¹) or hemangioma (F3¹), resulting in an increase of 1 to 2 orders of magnitude. From these results, it appears that the highest levels were observed in organs from mice F1¹, F2¹, and F5¹. Therefore, the levels of genomes in tumor-free organs appeared to be correlated to the levels in the tumors of the same mice.

TABLE 3.

Levels of genomes in tumor-free organs of tumor-prone and other mice^a

Litterb	Mouseb	Tumorc	Level of genomes in:
			Rib	Kidney	Skin	MG	SGf
Tumor prone	F11	MG	15d	5	7.2d	5	3.1d
	F21	MG	13.8	6.6	2.3	ND	<0.5
	F51	MG	NDe	>6d	>3.7d	ND	>1d
	F41	2 MG	3.7	4.1	<0.5	<0.5	ND
	F31	Hem	ND	3.6	ND	ND	<0.2
	F61	None	0.5	2.0	<0.3	<0.1	<0.1
Tumor free	Pool	None	1.1	0.3	<0.1	ND	0.5
	M4	None	ND	<0.1	ND	ND	ND
	M5	None	ND	0.5	ND	ND	0.35

^aSouthern blots of EcoRI-digested DNA (Fig. 7) from kidneys, tumor-free mammary glands, bone, skin, and salivary glands were sequentially hybridized with a PYV genomic probe and a probe for the cellular μj gene. The optical band intensity and the ratio of PYV to μj signal determined. 

^bMice are identified as previously from the L1 tumor-prone or from tumor-free litters. In the tumor-prone litter, three females (F1¹, F2¹, and F5¹) had one mammary gland tumor, one (F4¹) had two mammary gland tumors, and one (F6¹) was tumor free. For the pool, the organs of three mice sacrificed at 20 weeks postinfection were pooled. M4 and M5 represent two tumor-free independent males sacrificed at 34 weeks postinfection. 

^cThe presence of a tumor(s) in a given mouse is indicated. MG, mammary gland tumor; Hem, hemangioma; −, no tumor. 

^dThe presence of T1- or T3-specific defective-size genomes in the organs of mice F1¹ and F3¹, respectively, is indicated. 

^eND, not done. 

^fSG, salivary gland. 

DISCUSSION

Pattern of tumor development.

In the present year-long study of five litters of BALB/c mice infected at the neonatal stage with PYV, 3 of 18 males developed osteosarcomas and 5 of 16 females developed five mammary gland tumors and one hemangioma. This overall tumor frequency is normal for BALB/c mice. Similarly, the tumor target pattern is normal with respect to the tumor type and their apparent sex specificity (2, 5, 14). The osteosarcomas developed in three independent litters at 30, 37, and 57 weeks postinfection in an apparently random manner. In contrast, the mammary gland tumors and the hemangioma developed in five females belonging to the same litter. Furthermore, these tumors developed quasi-synchronously, becoming palpable between 28 and 31 weeks postinfection.

Most mouse strains, including BALB/c, are resistant to PYV-induced oncogenesis (3, 20, 21). The cytotoxic-T-cell-mediated antiviral immune response plays a major role in this resistance, as is the case for BALB/c mice (3, 21). Indeed, BALB/c nu/nu athymic mice are highly sensitive to PYV tumor induction (2, 14), and the transplantation of spleen cells from immunized wild-type BALB/c donors results in the elimination of mammary gland tumors in BALB/c nu/nu females (29). In resistant mice, rare tumors break through the immune response barrier. The nature of the events that lead to spontaneous breakthroughs has not been systematically investigated. The appearance of a tumor-prone litter was unexpected and suggests a biologically or accidentally caused litter-specific variation in the level of the immune response.

Biological hypotheses for such a litter-specific effect include the natural variations observed in murine gestation times. Pups that are dropped very early may be born with more abundant fetal tissues (hypothesis provided by an anonymous reviewer) and/or with a more immature immune system, both of which could alter the patterns of and the response to the acute and persistent phases of the infection. Further experiments are under way to test the hypothesis of an effect of the gestation time on tumor sensitivity.

Increase in the levels of viral genomes in all tissues in tumor-bearing mice and systemic spread of virus from the tumors to other organs.

In immunocompetent mice infected at the neonatal stage, the level of persisting genomes diminishes as a function of time (9, 31). This results in part from an age-related decrease in the capacity for viral replication in many tissues (31), in addition to the antibody-mediated virus clearance and cytotoxic T-lymphocyte-mediated elimination of virus-infected cells. In the present experiment, persistent viral genomes had reached the detection threshold for Southern blotting (<1 genome/cell) around 20 weeks postinfection (31). In contrast, in the tumor-bearing mice, the level of viral genomes in many organs was 1 to 2 orders of magnitude higher as late as 32 to 34 weeks postinfection. In comparisons of the genome load in organs of a given mouse, the number of genome copies per cell was by far the largest in the tumors, in the cases in which the tumors contained a replication-competent genome (F1¹, F2¹, F5¹, and M3²). The genome load organ pattern was different from that seen in de novo infection of adult mice, except for the bone (ribs), which had a high genome load as in de novo infection (31). Differences included a more extended range of organs in the tumor-bearing mice and organ-specific differences: for example, the genome load was higher in the pancreas (compared to de novo-infected mice) and lower in the tumor-free mammary glands. These differences may be related to the mode of viral dissemination, as discussed below, and are reminiscent of the alterations in the PYV distribution pattern among organs observed in function changes in the primary site of infection (9). The systemic increase in the levels of genomes in the organs of tumorous mice was most pronounced in three mice from the tumor-prone litter with mammary gland tumors; however, it was also detectable in a male mouse with an osteosarcoma (M3²). Therefore, it may be a general feature of PYV-induced oncogenesis following neonatal infection in BALB/c mice.

Two findings suggest that the viral genomes amplified in the tumor cells could have played a role in the systemic increase in viral genome load. First, in each tumorous mouse studied, there was an apparent correlation between the level of genomes in the tumor and in the nontumorous organs: the levels in the organs of mice F1¹, F2¹, F5¹, and M3² were higher than in those of mouse F4¹, in which the defective tumor-resident genome could not replicate. Second, in mice F1¹ and F5¹, where tumors T1 and T5 harbored a recognizable defective genome, the same tumor-resident genomes were the sole or dominant types in most of the 12 nontumorous organs tested. In contrast, the original wild type was present in only one or few organs, namely, the kidneys, and at much reduced levels.

Spread of capsid-defective genomes.

The deletions and rearrangements observed in the viral genomes of tumors T1 and T5 encompass the late region and would prevent the synthesis of functional capsid and the production of live virus. In particular, the T1 genome lacked the VP2/VP3 coding sequences and, in the tumor, the late-mRNA pattern was severely altered, both quantitatively and qualitatively. Such a defect would preclude a normal dissemination of viral genomes from the tumors to other tissues, i.e., by transport in encapsidated virus particles. Although wild-type virus was also present in these animals, it was present at much lower levels, mostly restricted to the kidneys, and was absent from the tumors. In tumor T1, the ratio of defective to putative wild-type helper was >50:1.

Two general categories of hypotheses may be entertained to explain the presence of the tumor-resident, capsid-defective genomes as the sole or major genome species in many nontumorous tissues. (i) A capsid-providing helper may play a role in establishing the defective viral genome in some or most tissues, prior to or at the initiation of tumorigenesis. The presumably wild-type helper virus is lost or eliminated from most tissues because of a selective advantage of the mutant genome, either in genome replication and/or in deleting a dominant epitope. A demonstration of wild-type genome persistence up to 34 weeks in the same 12 organs of 31 additional mice suggests that this is not a general phenomenon, at least in nontumorous animals (Fig. 7) (31).

(ii) Viral genomes can be disseminated from mammary gland tumors in the absence of normal capsids. The failure to detect live virus from the T1 tumor as well as from 12 organs in mouse F1¹ is congruent with, but does not prove, this hypothesis.

The present data cannot eliminate either hypothesis. However, they suggest that it might be important to consider the possibility of unconventional routes for genome dissemination. Precedents to support the hypothesis of dissemination and survival of capsidless PYV genomes and DNA can be found in the literature. A number of studies have examined the fate of purified naked PYV genomes or PYV-based plasmid DNAs introduced into mice by various routes. The introduction, intraperitoneally, subcutaneously, or by direct injection into the liver or spleen, of viral genomes or plasmids containing the complete genome results in efficient infection (8, 18). In this case, the relevance of this observation applies only to the first round of infection, since virus particles will thereafter be generated and disseminated in a normal manner. However, introduction of plasmids that do not contain the coding information for the capsid genes is also reported to lead to spread, replication, and temporary maintenance (4) or even transmission through the germ line (23). This is the case even when the plasmid contained only the origin of replication without the large T-antigen coding region (4). Thus, in the present era of DNA-mediated gene delivery resulting in systemic gene expression (32), the possibility of spread by nonencapsidated genomes needs to be considered. Since at least in the case of the AKR mouse strain, the mice in which tumors arise contain the highest levels of antiviral antibody (11), this mechanism would provide a route for evasion of the immune response. Since the infected mammary gland, despite its high potential for genome replication, produces very low yield of live virus (31) and since large segments of mammary gland tumors do not express viral capsids (28), this putative mechanism of spread could apply in the case of the wild-type genome as well.

In conclusion, the data presented here suggest that a significant systemic increase in the load of viral genomes took place in conjunction with mammary gland oncogenesis by PYV WTA2 in BALB/c mice infected at the neonatal stage. Furthermore, the data open the possibility that the viral spread may have been mediated by nonencapsidated viral genomes. This situation is reminiscent of the spread of human papillomavirus (HPV) in patients with primary HPV-positive cervical neoplasias. An increasing number of multiple primary HPV-positive cancers have recently been reported involving sites distal from the original tumor, e.g., lung and breast, reflecting the systemic spread of the HPV virus or genome (15, 16). Given the HPV episomal mode of replication and absence of late-gene expression in the basal skin epithelial cells, it may be important to consider that HPV spread may be mediated by the uncoated genome. This novel mechanism of spreading would provide an easy escape from the antiviral immune response and represent an important path in viral pathogenesis. Experiments are under way to further test this hypothesis.