The studies that we have performed during the past decade have allowed us to identify viral sequences responsible for the very high efficiency and restricted pathogenic potentential of MAV1-(N) [29–31]. We also established that MAV-induced nephroblastomas were polyclonal tumors  that constituted a unique model of the pediatric Wilms tumor .
The analysis of genomic libraries prepared from MAV1-induced tumors representing three different stages in tumor progression established that the MAV proviral genomes contained in the DNA of the tumor cells were not integrated in common sites. However, the relatively small size of the DNA insert transduced by the recombinant lambda phages, did not permit to exclude the possibility that MAV proviral genomes were inserted in common regions at the chromosome scale.
In order to determine whether MAV-induced rearrangements of the host genome were common to the three chicken nephroblastoma tumours that represented increasing developmental stages , we have isolated and characterized 78 BACs containing the normal DNA fragments corresponding to the insertional sites of MAV in the genome of these tumor cells. The molecular analysis of these BACs did not permit us to identify any common integration sites, but one, among the three different tumors.
It is well known that selective pressures likely occured during tumor progression and that the integration sites that we might identify at the late stages could be associated with events that led to tumor establishment. Therefore the lack of a common integration site, at a scale of 150 kb, probably reflected the various developmental stages and phenotypes of the tumors. On the other hand, our results suggested that preferential MAV integration sites might be conserved in the developed tumors, since independent junction fragments corresponding to different proviral genomes cloned from a given tumor, were hybridizing with the same BACs. These results suggested that the distribution of MAV integration sites in the tumors might not represent initial events but rather reflect the complex chromosome rearrangements that occur during tumor progression.
The use of BACs to perform a FISH analysis of the MAV integration sites permitted us to gain a better insight into the distribution of integration sites in the various tumors that we analyzed. In spite of the polyclonal nature of the nephroblastomas, a rather simple profile was obtained. The chicken genome is composed of 34 chromosomes, among which are 9 macrochromosomes and 25 microchromosomes. The MAV integration sites were found to be equally distributed between the micro and macrochromosomes that stained positive. However, these sites were not distributed randomly. Instead, the number of MAV integration site on chromosome 2 was much higher than expected and 3 integration sites were detected on this chromosome by two independent BACs. These finding suggested that during the establishment and progression of nephroblastomas, the maintenance of chromosome 2 alterations were preferentially selected. The results obtained by FISH confirmed that MAV proviral genomes were integrated in a limited number of sites, as previously predicted by junction fragment analysis  and pulse field electrophoresis .
During the preparation of this manuscript, Pajer et al. reported the use of inverse PCR and LTR-RACE to identify nephroblastoma-associated loci (Nals) in MAV2-induced avian nephroblastomas (Pajer, personal communication and manuscript in press). In order to compare the position of MAV insertion sites identified by FISH and PCR, we have calculated the physical localization of the Nals on each corresponding chromosome and found a fairly good match between the two sets of results (See additional file 1: Compilation of cytogenetic data obtained from FISH analysis). In both studies, MAV integration sites were found to be mainly distributed among chromosomes 1 and 2. Of particular interest was the identification of hot spots for proviral sequences at 1q1, and 2q2. Whether these sites represent preferential MAV integration sites or regions that contain genes required for tumor development remains an open question.
Two different types of information could be drawn from these observations: i) the distribution of MIRS and Nals points to chromosome regions that frequently harbour proviral MAV sequences in tumors; these regions likely contain genes that are important for tumor development; and ii) the reduced number of MAV integration sites that are maintained during tumor progression points to genes that are probably important at later stages, and the comparison of MAV integration sites in early and late tumors might help to distinguish between genes involved in the establishment and in the maintenance of the tumor state.
Based on the relatively well conserved synteny betwen man and chicken it was also possible to predict the nature of potential genes of interest. The use of normal and tumor RNAs as probes to identify BAC fragments that contain genes that are differentially expressed in normal and tumor tissue (figure 10) also provided critical information that could be used as another clue to assign potential genes to MAV insertion loci.
MAV integration sites were identified by FISH analysis (BACs 50, 65, 64, 97, 15) on Chromosome 2. Among the potential genes of interest contained in these areas Plag1 (8q12 in human), and twist (7p21) are target in 6% and 4% of MAV2-induced tumors (Pajer et al. In ress). Both encode transcription factors that are thought to play a role in tumorigenesis, LRCC (Leiomyomatosis and Renal Cell Cancer, 1q42-44 in human) is associated with papilloma renal cancer, while WTSL (Wilms tumor suppressor locus, 7p14-13 in human) is a potential suppressor gene whose alteration appears to be involved in normal kidney development and nephroblastoma. The CCN3/NOV gene (8q24-1 in human) which is a target for both MAV1 and MAV2-induced tumors (see below) is also localized on chromosome 2. The Gga 2 p3-2 zone represents a hot spot for MAV integration. Three BACs (85, 50 65) were mapped in this area. Bac 85 is a little more distal than the 3 others. BACs 50 and 65 are very close to each other but distinct. It is worth noting that Nal 2–28 (Pajer et al. in press) which include the twist gene overlaps with BACs 50 and 65. Our investigations also pointed out WTLS (at 7p11p15 in human) as a locus of interest for BACs 50 or 65.
Within the cytogenetic region Gga1q1 several genes of interest were potentially detected by FISH with BAC 22 and BAC 100. Among them, the cyclin-dependent kinase inhibitor 1B (CDKN 1B, 12p13 in human) and the N-ras oncogene (1p13 in human). In the same area, the data obtained by Pajer et al (in press) pointed to the POU2, OTF1, Oct1 transcription factors (1q22-3 in human). The localisation of BAC83 at Gga 1qter suggests as a potential target BIRC3 (Hsa 11q21) a candidate oncogene which is highly expressed in normal kidney and was reported to inhibit apoptosis.
At Gga3q24, the Wilms tumour 1 associated protein (WTAP, 6q25-27 in human) is a potential gene of interest for the region that is detected by two independant probes on BAC 71.
Two MAV1 insertion sites maped at 5q23-25 on the chicken genome. The human syntenic fragment (14q21-33), GPH (gephyrin at 14q23.3), TRAF3 (TNF receptor associated factor 3, at 14q32-33) and TGFβ3 (Transforming growth factor 3, at 14q24).
BACs 1 and 90 which maped at Gga 5 q23-25, contain MAV integration sites that were identified in two tumors representing different developmental stages. Both mapped very close to each other but did not co-localized since Bac1 is more proximal than Bac 90. The human syntenic fragment (14q21-33) contains several genes involved in kidney development and tumorigenesis. Among them, RCC2 (renal cell carcinoma 2, at 14q22-ter) is a locus which is lost in sporadic, non papillary renal cell carcinomas and oncocitomas. GPH (Gephyrin at 14q23.3) is a cytoplasmic, peripheral protein that anchors Gly-R. Although it is widely expressed, it is especially predominantly expressed in kidney. TRAF3 (TNF receptor associated factor 3 at 14q32-33) encodes an adapter protein that recruits other signaling molecules to the ligand-bound TNF family receptor. A gradient of TRAF3 is detected along the nephron, with progressive expression from proximal tubule to the collecting duct. TGFB3 is a well know transforming growth factor.
The region defined by BAC1 and 90 also corresponded to Nal 5–13 (Pajer et al. In press). Because these integration sites were identified in tumors representing different developmental stages, this area corresponded to a common integration region whose alteration is conserved during tumor progression, therefore suggesting that the gene(s) encoded by this portion of genome might be critical for nephroblastoma development and (or) tumor progression.
In addition to this situation, the two other integration sites identified in the most developed tumor by BACs 15 and 2 corresponded respectively to ccn3/nov (8q24.1 in human) and AdamTS1 (21q23.1 in human), two genes whose involvement in angiogenesis, matrix remodeling and tumorigenesis is well documented [7, 19–27]. The ccn3 gene was previously mapped on chicken chromosome 2q34-36 . Although the present study, and the results of Pajer et al. indicated that ccn3 is not a common integration site for MAV, this gene was identified as a MAV target in both studies. However, the MAV2-induced tumors analyzed by Pajer et al. did not show any increase in ccn3 expression.
Since both the MAV1- and the MAV2-induced nephroblastomas that we analyzed showed elevated levels of ccn3 expression , these conflicting observations result from either the route of injection, the time frame for injection, the different nature of the viral strains or host differences. The MAV2 (O) strain that was used in our previous studies was molecularly cloned and sequenced . It induced 20% nephroblastomas, as opposed to 100% efficiciency of the MAV1(N) strain. In both cases, nephroblastomas were induced after intraveinous injection of 14 day-old embryos or intraperitoneal injection of day old chicken . Since we have established that blastemal cells undergoing epithelial differentiation are the targets for MAV1, the time frame and route of injection may be critical. Indeed, the blastemal cells express high levels of ccn3 (Cherel et al. manuscript in preparation). Therefore, the elevated levels of ccn3 expression detected in all MAV-induced nephroblastomas might result from the expansion of blastemal cells that are transformed at a well defined stage of differentiation upon MAV1 infection.
Hybridization of BACs containing MAV1-integration sites with labeled mRNAs isolated from normal kidney tissue and nephroblastomas also permitted us to perform an analysis of the genes that are proximal to MAV integration sites and that are differentially expressed in normal and tumor condition. Among the different genes that were uncovered in this study, the ADAMTS1 gene was of potential interest. The ADAMTS1 protein is a matricellular proteinase known to participate in the late stages of tumorigenesis. Forty-five percent of newborn ADAMTS1 null mice died, probably as a result of kidney malformation that becomes apparent at birth .
Comparison of the expression pattern of CCN3 and ADAMTS1 shows striking similarities. In both cases overexpression of the protein is detected in all tumors tested, while the MAV proviral sequences are detected only once in the vicinity of these genes. These observations suggest that MAV -induced nephroblastoma occurs via a multistep process that involves a cascade of proteins acting along a common signaling pathway. Direct or indirect alteration of any step could result from MAV integration within or in the vicinity of critical genes whose increased expression would eventually be required for tumor progression. The identification of TGFβ 3 locus as a target for MAV integration in two independent tumors (501 and 725) is in favor of such an hypothesis. The role of TGFβ 1 in expression of CCN genes expression has been widely documented and the antagonistic activity of TGFβ 1 and TGFβ 3 has been shown to be critical in several instances. The activation of TGFβ3 expression by MAV might therefore result in an increased expression of CCN3 in tumors, similar to that observed upon integration of MAV within the ccn3 gene itself. Interestingly, tumor 725 which is the most developed is the only one in which integration of MAV occured in three gene loci whose alterations would have cummulative effects. A less developed tumor such as 501 only shows integration in the vicinity of TGFβ 3 and the early developed tumor does not show any of them.
In summary, our present study suggests that the development of nephroblastoma from an initial diffuse tumor phenotype (501D) to a well developed compact tumor (725) is accompanied by the selection of MAV integration sites in chromosome loci where genes involved in kidney differentiation are localized. The alteration of any of these genes by MAV integration at early stages of blastemal cell differentiation, would trigger the tumorigenic process. The multiplicity of potential genetic and cellular targets would provide support to the very high efficiency of MAV1 (N) which can induce 100% nephroblastomas within a 8-week period of time post injection. It will be interesting to determine whether the phenotypic variability of the MAV-induced nephroblastomas compares to the Wilms' tumors situation, and if the various subtypes of tumors result from different sequences of genes alterations.