The MET Gene Is a Common Integration Target in Avian Leukosis Virus Subgroup J-Induced Chicken Hemangiomas

ABSTRACT

Avian leukosis virus subgroup J (ALV-J) is a simple retrovirus that can cause hemangiomas and myeloid tumors in chickens and is currently a major economic problem in Asia. Here we characterize ALV-J strain PDRC-59831, a newly studied U.S. isolate of ALV-J. Five-day-old chicken embryos were infected with this virus, and the chickens developed myeloid leukosis and hemangiomas within 2 months after hatching. To investigate the mechanism of pathogenesis, we employed high-throughput sequencing to analyze proviral integration sites in these tumors. We found expanded clones with integrations in the MET gene in two of the five hemangiomas studied. This integration locus was not seen in previous work characterizing ALV-J-induced myeloid leukosis. MET is a known proto-oncogene that acts through a diverse set of signaling pathways and is involved in many neoplasms. We show that tumors harboring MET integrations exhibit strong overexpression of MET mRNA.

IMPORTANCE These data suggest that ALV-J induces oncogenesis by insertional mutagenesis, and integrations in the MET oncogene can drive the overexpression of MET and contribute to the development of hemangiomas.

INTRODUCTION

Avian leukosis viruses (ALVs) are classified into subgroups based on their envelope gp85 surface glycoprotein (SU), viral cross-neutralization patterns, and host range. Most avian retroviruses are classified as subgroup A, B, C, D, or E. ALV subgroup J (ALV-J) was first isolated in 1988 in the United Kingdom; the prototype strain, HPRS-103, causes primarily myeloid leukosis but can induce other tumor types at low incidence (1–3). ALV-J is thought to have originated from a recombination event between an exogenous ALV and an ancient endogenous avian (EAV) retroviral element (4, 5). This recombination event incorporated the endogenous retroviral env into ALV-J.

Since its discovery in the United Kingdom, a variety of ALV-J strains in diverse geographical areas, including North America, Europe, East Asia, Australia, and the Middle East, have been characterized (6–11). It is believed that these isolates derive from a single common ancestor and are not the result of independent recombination events (6). The types of neoplasms caused by ALV-J vary and can be influenced by the specific strain of ALV-J that has infected the bird. Most often, the virus induces tumors of myeloid origin (as with HPRS-103), but some strains induce primarily hemangiomas. Hemangiomas are vascular tumors found in the skin or visceral organs that originate from endothelial cells that line blood vessels (12). Other ALV-J strains are capable of inducing both myeloid tumors and hemangiomas. We found that the strain used in this study falls into the third category, inducing both hemangiomas and myeloid tumors at a high incidence.

Besides myeloid tumors and hemangiomas, ALV-J has been shown to induce other types of tumors at a low frequency, such as skeletal myelocytomas, renal tumors, histiocytic sarcomas, and others (2, 13). This pathology contrasts sharply with that of the more studied ALV-A, which induces mainly B-cell lymphomas but also erythroblastomas (14–16).

ALV-J infection can cause significant economic losses due to reduced egg production, stunted growth, and early death. The economic losses have been particularly extensive in China, where the virus commonly infects poultry (17). It was recently shown that ALV-J infection is not limited to domesticated chickens. In fact, infection with ALV subgroups A, B, and J appears to be widespread in wild fowl throughout China (18, 19).

ALVs do not carry a viral oncogene and instead cause neoplasia through insertional mutagenesis (20). In order to complete the viral life cycle, all retroviruses must integrate into the genomic DNA of the infected cell. Thus, the provirus can act as a mutagen, landing within a gene and ablating its function. Alternatively, because the virus has potent enhancers and promoters in its long terminal repeats (LTRs), ALV can induce the expression of genes located near an insertion site. This process can drive tumor formation if the provirus integrates near and perturbs the expression of cancer-related genes.

Virus-induced mutagenesis can be exploited to identify genes that may play a role in driving the development of neoplasms. For example, the virus can be used to induce tumor formation, and common integration loci can be identified. These common integration sites (CISs) flag a genomic locus as potentially harboring an oncogene or tumor suppressor. Viral insertional mutagenesis screens have been fruitful in identifying cancer genes in several model systems (20). ALV is an especially useful virus for such a screen because it integrates in a largely random fashion, with only a slight preference for active transcriptional units (21, 22). This ensures that as many genomic loci as possible are probed for oncogenic potential by the virus.

Previous studies implicated several genes as drivers of tumorigenesis in ALV-induced neoplasms. In ALV-A-induced B-cell lymphomas, common integration sites were identified near or within the MYC, MYB, MIR-155, and TERT genes (23–26). The study of proviral integrations within ALV-J-induced neoplasms has only recently begun. Early work showed MYC, TERT, and ZIC1 to be targets of proviral integration in ALV-J-induced myeloid leukosis (ML) (27).

In this study, we conducted an insertional mutagenesis screen to identify the genes involved in ALV-J-induced tumors. To identify viral integration sites in these tumors, we employed high-throughput sequencing on the Illumina platform. We identified intron 1 of the MET gene as a common integration site in hemangiomas. MET encodes a well-studied receptor tyrosine kinase that binds hepatocyte growth factor/scatter factor (HGF/SF) and plays important roles in normal development and a wide range of human cancers (28). Because we observed integrations near a known oncogene in multiple tumors, we hypothesize that ALV-J tumors, like those induced by ALV-A, are generated by insertional mutagenesis.

MATERIALS AND METHODS

Tumor induction.

ALV-J strain PDRC-59831 was isolated from a 38-week-old broiler breeder chicken. The case was recorded on 30 May 2007 on a farm near Danielsville, GA. In this study, ALV-J strain PDRC-59831 was inoculated into 30 5-day-old SPAFAS embryos (Charles River) via the yolk sac route. Four embryos died at day 10 of embryogenesis, leaving 26 viable embryos. Of these embryos, 11 hatched (11/26; 42%), 5 pipped but could not complete hatching, and 10 did not hatch and died. In comparison, 90% (19/21) of the uninoculated control chickens hatched. Chickens were observed daily and euthanized when apparently ill or at 12 weeks of age. Of the 11 ALV-J-infected chickens, one was euthanized at week 2 and three died at week 11 for reasons unrelated to infection. One chicken died at week 5, and one was euthanized at week 7; both had tumors. Five chickens were euthanized at 12 weeks of age, all of which had tumors. In total, tumor tissue was obtained from six ALV-infected chickens. The early death of bird 7 (B7) prevented the collection of useful tissue. Tumors were classified by gross examination, and tissue samples were collected, flash-frozen, and then stored at −80°C.

Phylogenetic analysis of the ALV-J isolate PDRC-59831 env gene.

Phylogenetic analysis was conducted by using the neighbor-joining method (29). The percentages of replicate trees in which the associated taxa clustered together in the bootstrap test (500 replicates) are shown next to the branches (30). The evolutionary distances were computed by using the maximum composite likelihood method (31) and are in units of the number of base substitutions per site. The analysis involved 16 nucleotide sequences. GenBank accession numbers are as follows: HQ634811 for HUB09JY03, Z46390 for HPRS-103, AF307951 for UD4, AF307592 for UD5, KP284572 for PDRC-59831, HM235670 for YZ9902, DQ115805 for NX0101, HM235669 for NM2002-1, HM235667 for JS-nt, FJ216405 for SD07LK1, HQ425636 for SCDY1, HM235668 for NHH, JN624878 for JL093-1, GU982310 for JS09GY6, GU982308 for JS09GY3, and AY027920 for ADOL-7501. All positions containing gaps and missing data were eliminated. These analyses were conducted with MEGA6 (32).

DNA extraction and deep sequencing.

Fifty milligrams of tissue was homogenized with a Kimble-Chase Kontes pellet pestle and then digested with proteinase K at 50°C for 15 h. The sample was then phenol extracted, put through a 25-gauge needle 10 times, and ethanol precipitated; this procedure was repeated. The sample was then treated with 2 μg RNase A for 1 h at 37°C. Phenol extraction, shearing, and ethanol precipitation were repeated, and the DNA concentration was measured with a Thermo Scientific Nanodrop 2000c instrument. Five nanograms of purified genomic DNA was sonicated with a Bioruptor UCD-200 instrument. End repair, A tailing, and adapter ligation were performed as described previously by Gillet et al. (33) (adapter short arm, P-GATCGGAAGAGCAAAAAAAAAAAAAAAA; adapter long arm, CAAGCAGAAGACGGCATACGAGATXXXXXXGTGACTGGAGTTCAGACGTGTGCTCTTCCGATC*T [where the X's denote the barcode sequence, P denotes phosphorylation, and * denotes a phosphorothioate bond]). Multiplexed barcodes each differed by at least 2 nucleotides. Two rounds of PCR (nested) were conducted to enrich the library for proviral junctions. The first PCR had 23 cycles and employed one ALV-J-specific primer (GGGACTGTAGCATGTATAGGCGCTGAG) between the 3′ LTR and env and a second primer within the adapter which was ligated previously (CAAGCAGAAGACGGCATACGAGAT). The second round of PCR employed a primer (AATGATACGGCGACCACCGAGATCTACACTCGACGATTGCGAGCACCTGAATGAAG) at the 3′ end of the LTR, 12 nucleotides short of the junction between virus and genomic DNAs, as well as a nested adapter primer overlapping the barcode sequence within the adapter (CAAGCAGAAGACGGCATACGAGATXXXXXX). Libraries were quantified by quantitative PCR (qPCR) and then underwent single-end 100-bp multiplexed sequencing on the Illumina Hi-Seq 2000 platform. A custom sequencing primer (ACGATTGCGAGCACCTGAATGAAGTGAAAGG) was used, which hybridized near the end of the viral 3′ LTR, 5 nucleotides short of the junction between viral and genomic DNAs. This allowed for reads to be validated as genuine integrations, by verifying that each read begins with the last 5 nucleotides of the proviral DNA. In all, 9,759,304 junction reads were obtained.

Sequence analysis.

Reads were first filtered with a custom python script to remove sequences that did not begin with the last 5 nucleotides of viral DNA, “CTTCA.” A total of 1,352,053 reads (13.85%) were discarded because they did not begin with this verification sequence. The files were then uploaded to Galaxy (34–36), which was used to perform downstream analyses. In Galaxy, the quality scores were first converted to Sanger format with FastQ Groomer v1.0.4 (37). Adapters were then trimmed by using Galaxy Clip tool v1.0.1. This tool also removed reads containing an N (14,871; 0.18% of reads removed at this step) and reads <20 nucleotides long after adapter removal (4,289,006; 51.02%). The remaining reads (4,103,370; 48.81%), were mapped with Bowtie (38), using the Gallus gallus 4.0 genome (November 2011). A total of 100,000 random mapped reads were selected from each sample to be used for further analysis. If <100,000 reads were present for a sample, all available reads were used. A custom Perl pipeline was developed to analyze the aligned-read output from Bowtie. Briefly, reads containing sequencing errors were filtered, and read counts and breakpoints were quantified. Integrations found in multiple samples were assigned to the sample with the highest number of breakpoints. Integration hot spots were identified by genome walking. If two integrations were within 5 kb of each other, they were placed into an integration group. Additional integrations were included in the group until an integration-free span of 5 kb was reached. Files were annotated with RefSeq features, and the orientation and distance to the nearest gene were calculated for each integration. The source code is available upon request.

MET gene expression.

RNA transcripts were isolated from all five hemangiomas, three myeloid tumors, and four nontumor controls with RNA-Bee (Tel-Test). First-strand synthesis was performed with a poly(dT) primer and Superscript III reverse transcriptase (Invitrogen) according to the manufacturer's protocol. Quantitative reverse transcription-PCR (qRT-PCR) was conducted with iQ SYBR green Supermix (Bio-Rad) according to the manufacturer's protocol. MET primers (AGGACATTTTGGGTGTGTGT and AACTGAGCCACTTCTTCCAG) were designed by using primer3plus software (http://www.primer3plus.com/). Thermal cycling was conducted on a Bio-Rad C1000 thermal cycler/CFX96 real-time system. MET expression levels were normalized to GAPDH expression levels by using primers reported previously (39). PCR was repeated four times, and each sample was present in duplicate during each run. Results were normalized to B3 kidney hemangioma values by using the comparative threshold cycle (C_T) method. Melt curves were conducted to verify the specificity of the primers.

Nucleotide sequence accession number.

The PDRC-59831 consensus viral genome sequence has been deposited in NCBI GenBank under accession number KP284572.

RESULTS

Characterization of ALV-J isolate PDRC-59831.

The tumors in this study were induced by ALV-J strain PDRC-59831, an American isolate of ALV-J that has not been previously described in the literature. Strain PDRC-59831 was isolated in 2007 from a meat-type chicken. The provirus was isolated from infected cells by PCR, four isolates were sequenced, and the consensus sequence was submitted to GenBank (accession number KP284572). The env gene sequence was compared to env genes from other strains of ALV-J isolated from around the world. Phylogenetic analysis of the PDRC-59831 env gene (Fig. 1) shows a close relationship with those of two other American ALV-J isolates, ALV-UD4 (96.2% similarity) and ALV-UD5 (96.3%), as well as that of Chinese isolate YZ9902 (95.3%). These isolates, along with PDRC-59831, cluster with the original United Kingdom isolate HPRS-103 (95.0% similarity).

FIG 1.

Sequence comparison and phylogenetic analysis of env genes of ALV-J isolates. Phylogenetic analysis was conducted by using the neighbor-joining method. The optimal tree, with a sum of branch length of 0.37621, is shown. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (500 replicates) are shown next to the branches. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. Strain PDRC-59831 (in boldface type) is shown to cluster closest to UD4, UD5, and YZ9902 and on the same branch as prototype ALV-J isolate HPRS-103.

Several studies have reported certain genetic alterations in ALV-J strains that have a propensity for generating hemangiomas. For example, an 11-nucleotide deletion was observed in the U3 region of the LTR of ALV-J strains SCDY1 and NHH (40). Also, two different 19-nucleotide insertions, one in the 5′ untranslated region (UTR) and another in U3, were identified in hemangioma-inducing strains JL093-1, SD09DP03, and HLJ09MDJ-1 (41). Sequence alignments were carried out to determine if PDRC-59831 contained any of these alterations, and it does not. Instead, it shares similarity with the ML-inducing prototype strain HPRS-103 in these regions (data not shown). This finding suggests that these genetic alterations are not required for hemangioma induction, but it does not rule out their possible contribution. Recently, a 205-nucleotide deletion in the 3′ UTR has been identified in some strains of ALV-J, which leads to higher oncogenicity and lethality rates in infected chickens (42). We determined by alignment that PDRC-59831 contains a 216-nucleotide deletion in this region. This deletion overlaps 187 of the 205 nucleotides described previously.

Sequencing of ALV-J/cellular integration junctions.

All 7 birds infected with PDRC-59831 developed tumors (hemangiomas, myeloid tumors, or both) by 12 weeks of age. Six hemangiomas and four myeloid tumors were observed in total (Table 1). Genomic DNA was isolated from these tumors and randomly sheared by sonication, and proviral junctions were amplified in two successive rounds of nested PCR, as outlined in the legend of Fig. 2. Proviral integration/genomic DNA junctions were deep sequenced on the Illumina Hi-Seq 2000 platform and mapped onto the Gallus gallus genome.

TABLE 1.

Tissues collected and neoplasms observed^a

Bird	Neoplasm observed in tissue
	Bursa	Liver	Spleen	Brain	Kidney	Breast	Leg	Intestine	Lung	Head
B1	X	X	X	X	−	M	H	−	−	H
B2	X	X	X	X	−	−	−	−	H	−
B3	X	X	X	X	H	−	−	−	−	−
B4	X	X	X	X	−	−	−	H	−	−
B5	X	M	X	X	−	−	−	−	−	−
B6	X	X	X	X	M	−	−	−	−	−
B7	X	H	X	X	M	−	−	−	−	−
C1	X	X	X	X	−	−	−	−	−	−
C2	X	X	X	X	X	X	X	−	−	−

^aX, no neoplasm; H, hemangioma; M, myeloid tumor. −, tissue was not collected, and no neoplasm was observed during dissection. The tissues from bird 7 (B7) were not analyzed in this study.

FIG 2.

Library preparation. Genomic DNA was isolated from tumor and nontumor samples. DNA was sonicated, end repaired, and A tailed. Adapters containing barcodes were ligated onto the DNA fragments, and they underwent two rounds of PCR (nested). The final product underwent multiplexed sequencing on the Illumina Hi-Seq 2000 platform.

We chose to sonicate the DNA to induce random fragmentation. This introduces different sonication breakpoints into unique DNA molecules across multiple cells. Sometimes, multiple cells in a sample carry the same viral integration site due to clonal expansion. When this is the case, shearing the DNA from these cells can produce multiple fragments that share the same integration site but have different sonication breakpoints. By quantifying these breakpoints after deep sequencing, we were able to determine the relative abundance of an integration with respect to the other integrations in the sample. We refer to integrations that exhibit more than one sonication breakpoint as “expanded clones” (Fig. 3A and 4).

FIG 3.

Distribution of integration sites in ALV-J-induced hemangiomas. (A) A total of 30,850 breakpoints were identified in ALV-J-induced hemangiomas. Genes that had 10 or more breakpoints are highlighted in a separate pie. Each slice represents a unique integration site, and the size of the slice represents the number of breakpoints for that site. The table shows integrations with the highest number of breakpoints along with the tumor in which that integration was identified. ncRNA, noncoding RNA. (B) The top 10 most abundant integrations identified in head and leg hemangiomas are shown. Integrations in the MET gene represent the most abundant integration sites in these tumors.

FIG 4.

ALV-J integration in the MET gene in chicken hemangiomas and nontumor tissues. There were 9 integration sites at the MET locus in 5 hemangiomas (top) as well as 10 integration sites in 24 nontumor tissues (bottom) from infected birds. No MET integrations were observed in myeloid tumors. Integrations where multiple breakpoints were observed are represented as solid arrows, and the corresponding number of breakpoints is noted. Arrowheads represent the genomic location and the transcriptional orientation of each provirus. Most of the multibreakpoint integrations were observed in the intron 1 region of the MET gene.

A total of 25,961 integrations were identified among five ALV-J-induced hemangiomas analyzed. Integrations were mapped and quantified with a custom data analysis workflow (see Materials and Methods). A total of 15,738 integration “hot spots” were identified among all myeloid tumor, hemangioma, and nontumor tissues. These hot spots were defined as any genomic loci containing at least two integrations where each integration is separated by no more than 5 kb. The top 20 hot spots are shown in Table 2. Although these loci are sites of frequent integration, breakpoint analysis showed that these integrations are not found in highly expanded clones.

TABLE 2.

Top 20 common integration sites in ALV-J-induced hemangiomas^a

Gene	No. of common integrations	Avg no. of breakpoints
LOC101750146	29	1.0
ELF1	29	1.3
LAPTM4A/WDR35	19	1.0
MAML2	17	1.1
SPRY1/ANKRD50	17	1.5
NFKBIA	17	1.1
MAP4K4	16	1.1
ADARB1/GLS/STAT1	16	1.2
SSBP3	15	1.1
GSTA3/ICK	15	1.1
ATAD2/FBXO32	15	1.2
ADD3	14	1.4
RUNX3	14	1.1
SPIDR	14	1.1
RREB1	14	1.1
SRGAP3	14	1.2
INPP5A	13	1.2
ANKRD9	13	1.1
TNFRS21	13	1.0
GRHL1	13	1.0
MET	3	16.0

^aHot spots were defined as any genomic locus containing at least two integrations within 5 kb of each other. By definition, hot spots are flanked on either side by at least 5 kb of sequence lacking an integration. The number of integrations observed in hemangiomas and the average number of breakpoints are listed for each hot spot. MET is shown for comparison.

Interestingly, the endogenous retrovirus gene LOC101750146 is tied for the most frequent target of integration among the hemangiomas that we analyzed, with a total of 29 integrations (Table 2). However, none of these clones were highly expanded, and many integrations were also seen in nontumor tissues, so these integrations are unlikely to be relevant to tumor induction.

Some ALV-J integrations are present in expanded clones.

In order to measure the relative abundance of each integration, we quantified the number of sonication breakpoints for each site. Previous studies have shown that sonication breakpoints can be used as a measure of clonal expansion; if an integration has more than one breakpoint, the cell carrying that integration has undergone clonal expansion (43).

A total of 88.5% of the integrations that we identified exhibited only a single sonication breakpoint, suggesting that these integrations were not clonally expanded, while 11.5% had two or more breakpoints, evidence of clonal expansion. The 17 most highly expanded clones are shown in Fig. 3A. Each of these integrations had 10 or more breakpoints. The tumor that we refer to as a “head hemangioma” is a hemangioma that developed subcutaneously outside the skull.

MET intron 1 is an expanded common integration site.

Breakpoint analysis revealed three multibreakpoint integrations within the first intron of the MET gene. These integrations occurred in a head hemangioma (2 integrations) and a leg hemangioma. Strikingly, these MET integrations represented the most highly expanded clones in both tumors (Fig. 3B). To confirm the deep sequencing results, genomic DNA/proviral junctions from this locus were PCR amplified and sequenced. This verified that these integrations were present only in the expected tumors and in the orientation and location predicted by deep sequencing (data not shown). Interestingly, the two MET integrations in the head hemangioma occurred in different orientations and were determined by deep sequencing (of the 3′ ends) to be offset by only 6 nucleotides. We were able to verify the 3′ end of each of the proviral junctions by PCR and sequencing, but were unable to amplify the 5′ junction in both cases. This suggests that these two proviruses are adjacent to each other on the same strand and in opposite orientations. The 6-nucleotide offset observed by deep sequencing could be the result of the characteristic 6-nucleotide duplication that occurs upon ALV integration (44). Several lower-breakpoint MET integrations were observed in infected nontumor controls (Fig. 4), but none were seen in myeloid tumors.

All three multibreakpoint integrations identified in MET hemangiomas existed in a tight cluster within intron 1, a region spanning 1,599 bp. This suggests that ALV-J integration within this specific region of MET gave these tumor cells a selective advantage and may have driven hemangiomagenesis in these birds. Both of these tumors had at least one highly abundant integration in the same orientation as the MET gene. Since the first exon of MET is noncoding, this integration pattern suggests that the virus may be inducing the overexpression of the MET protein.

MET mRNA expression is elevated in tumors containing MET proviral integrations.

Next, we wanted to determine the effects of MET intron 1 integrations on gene expression in these tumors. ALV LTRs are known to harbor strong promoter and enhancer sequences that promote gene expression near viral integration sites (20). MET mRNA expression was analyzed by qRT-PCR (Fig. 5). As expected, we found that MET mRNA expression correlated closely with the abundance of the MET integration sites in the sample, as measured by breakpoint analysis. For example, the head hemangioma, which has two integrations (21 and 23 breakpoints), exhibited the highest MET expression level of any tumor. The expression level was a full 131-fold higher than that in the kidney hemangioma, which lacked any high-abundance MET integrations. The same is true for hemangioma of the leg, which also contained a MET expanded clone and showed higher levels of MET mRNA expression than the samples that lacked MET integrations. This finding suggests that MET expression is induced by viral integration and plays a causal role in hemangioma development.

FIG 5.

MET mRNA expression measured by qRT-PCR. Hemangiomas (H) and myeloid tumors (M) as well as nontumor and uninfected controls (labeled control) are shown. The MET expression level was normalized to that of GAPDH and is relative to that of the hemangioma with the lowest MET expression level, B3 kidney hemangioma. Error bars represent standard errors of the means. Tumors containing highly abundant MET integrations are noted (asterisks). These tumors exhibited markedly increased levels of MET mRNA expression compared to those in other tumors and nontumor controls.

DISCUSSION

In this study, we present the first integration profiles of ALV-J-induced hemangiomas. Hemangiomas are vascular tumors characterized by uncontrolled angiogenesis (45). They can occur in many species, including humans. In fact, infantile hemangioma is the most common tumor of early childhood, estimated to develop in 7 to 10% of infants (46). Hemangiomas present as benign tumors that grow postnatally for 8 to 12 months and then typically undergo a slow process of self-involution that can last several years (47, 48).

Previous studies analyzing ALV-induced neoplasms focused mostly on ALV-A-induced B-cell lymphoma and identified MYC, MYB, MIR-155, and TERT as common integration sites in these tumors (23–26). Only recently have ALV-J-induced neoplasms been the subject of similar work. A recent study of ALV-J-induced myeloid leukosis showed that MYC, TERT, and ZIC1 are targets of proviral integration in myeloid tumors (27). Interestingly, we did not identify these genes as CISs in this study. This may be due to the fact that only three myeloid tumors were analyzed. It is conceivable that with a larger cohort of tumors, these CISs may have been observed as well.

In this study, we implicate MET overexpression as a causal agent in the development of ALV-J-induced hemangiomas in chickens. Interestingly, although virus-induced tumors have been the subject of much study, to the best of our knowledge, no previous work has implicated MET in tumor induction by insertional mutagenesis. This suggests that MET may play a unique role in inducing hemangiomas rather than other types of virus-induced tumors.

MET has been studied extensively and has important roles in both development and cancer. The MET protein is a receptor tyrosine kinase that binds hepatocyte growth factor/scatter factor (HGF/SF) and can activate an array of downstream signaling pathways, including phosphatidylinositol 3-kinase (PI3K)-AKT, RAC1-CDC42, RAP1, and RAS–mitogen-activated protein kinase (MAPK) (49). Previous work has shown that MET is activated in many types of human cancer via mutation, amplification, and protein overexpression, and MET activation correlates with poor prognosis in cancer patients (50–52).

It has also been shown that MET plays an important role in angiogenesis, a process that is crucial to the development of hemangioma. For example, activation of the HGF/SF-MET pathway is now understood to be a potent inducer of angiogenesis, specifically in endothelial cells, the same type of cells that give rise to hemangiomas (50, 53–55). In addition, the HGF/SF-MET pathway can suppress TSP1, a negative regulator of angiogenesis, and can induce the expression of VEGFA, a proangiogenic gene. In this way, the HGF/SF-MET pathway controls an “angiogenic switch,” turning on angiogenesis (56). Additionally, it has been shown that activating mutations in MET can induce hemangiosarcomas in mice (57). Viewed in this context, it does not seem surprising that ALV-J-induced MET overexpression can lead to hemangiomagenesis in chicken. Whether or not MET can act as a causal agent in human hemangioma has yet to be determined, but this may be an avenue for further research.

In addition to MET, other CISs were identified as hot spots of integration (Table 2). Notably, none of these integrations were seen in highly expanded clones, and they were also observed in nontumor tissue. This finding suggests that these CISs are unlikely to be relevant to tumor induction. In contrast, some tumors exhibited non-MET integrations in highly expanded clones. For example, TRIO and EYA4 both contained integrations with many breakpoints. TRIO has guanyl exchange factor (GEF) activity and regulates Rho family GTPases, which coordinate cytoskeletal rearrangement and cell migration (58, 59). EYA4 is a member of the eyes absent (EYA) family of proteins; it has phosphatase activity and may function in eye development as a transcriptional activator (60). Our data suggest that these genes may also be capable of contributing to hemangioma formation, but because similar integrations were not observed in multiple tumors, we cannot establish with certainty that they are drivers of oncogenesis and not merely passenger integrations. Further studies involving a larger cohort of birds may help identify other common integration sites and genes that drive ALV-J-induced hemangiomagenesis.