Misexpression of Full-length HMGA2 Induces Benign Mesenchymal Tumors in Mice

HMGA2 is a member of the high-mobility group AT-hook (HMGA) family proteins that bind in the minor groove of the AT-rich regions of DNA through three conserved DNA-binding domains, termed AT-hooks (1). Although devoid of ability to transcriptionally activate promoters by themselves, these proteins play an important role in the transcription at specific target promoters by not only increasing the DNA-binding affinity of transcription factors via modulation of the local DNA architecture but also facilitate the assembly and disassembly of multiprotein complexes, called enhanceosomes, that are recognized by RNA polymerase II with higher efficiency (2). A well-characterized example is for HMGA2, which has been shown to bind to and modulate the activities of the promoters of cyclin A and ERCC1 genes (3, 4).

HMGA2 protein plays a critical role in development, where it is almost exclusively expressed in the undifferentiated mesenchyme (5). Accordingly, its expression is largely restricted to the embryonic stage and decreases to undetectable levels in most adult differentiated tissues (6, 7). Consistent with its role in mesenchymal proliferation and differentiation, Hmga2-null mice exhibit a pygmy phenotype. These mice weigh only 40% of wild-type littermates due to a reduction in the mesenchymal cell number but not cell size (7). Also, Hmga2-null embryonic fibroblasts show 4-fold fewer cells in culture, indicating a defect in cellular proliferation (7). The importance of the role of HMGA2 in mesenchymal proliferation and differentiation is exemplified by the results of a study that showed that Hmga2-null and heterozygous mice are resistant to both diet-induced and genetic obesity due to the retarded proliferative capacity of mesenchymally derived preadipocytes (8).

The role of HMGA2 in mesenchymal proliferation and transformation was further substantiated when it was shown that this gene is disrupted by translocation at 12q13-15 in a number of human mesenchymal tumor types (9). These tumors include lipomas (9, 10), pulmonary chondroid hamartomas (11), uterine leiomyomas (12), fibroadenomas of the breast (13), salivary gland adenomas (14–16), adenolipomas of the breast (17), and endometrial polyps (18, 19). The chromosomal translocations result in derepression of the HMGA2 gene, and it is hypothesized that misexpression of HMGA2 in a differentiated mesenchymal cell brings about the tumorigenic change (20). The nature of the chromosomal disruptions of the HMGA2 gene is thought to be an important determinant of the transforming capability of the resulting aberrant transcripts, which are frequently truncated immediately after the third AT-hook domain. In some cases, chimeric proteins result from the fusion transcripts that contain transactivation sequences, such as the LIM domains or serine-threonine–rich acidic domains (9–11, 21). Therefore, it has been postulated that the loss of the 3′-end of the gene is necessary for the tumorigenic pathway (22, 23). However, this hypothesis did not explain a subset of the same tumor types with translocations involving chromosome 12 that have been found to retain the full coding region of HMGA2 (12, 24, 25), suggesting that the nature of the transcript may not be important.

We attempted to make transgenic mouse models to address the specific question whether misexpression of full-length, truncated, or both versions of the HMGA2 transcripts under a differentiated-mesenchyme-specific promoter would give rise to mesenchymal tumors. Therefore, we expressed full-length and truncated human HMGA2 transcripts in transgenic mice under the control of the differentiated-adipocyte-specific promoter of the adipocyte P2 (aP2/Fabp4) gene. The transgene produced HMGA2 protein in a wide range of tissues and the transgenic mice showed a neoplastic phenotype, including fibroadenomas of the breast and salivary gland adenomas. These results provide the first in vivo evidence that HMGA2 misexpression in differentiated mesenchyme has a causal role in producing mesenchymal tumors, and that the truncation of HMGA2 and/or addition of ectopic fusion sequences are not necessary to the ability of HMGA2 to produce these types of neoplasia.

Transgene constructs and mice. The 330 bp cDNA of human HMGA2 open reading frame (ORF) was amplified by reverse transcription-PCR (RT-PCR) of total RNA from DLD-1 colorectal cancer cell line using a forward primer 5′-AATCTAGAGATGAGCGCACGC-3′ and a reverse primer 5′-AACCATGGCGCCCCCTAGTCCTCTTCGG-3′. To direct the transgenic expression of HMGA2 to adipose tissue, the full-length human HMGA2 ORF was placed under the control of 5.4 kb aP2 promoter fragment (26). To enhance the expression of the transgene, a genomic fragment of human β-globin gene was cloned downstream of the HMGA2 cDNA. This 1.7 kb BamHI/PstI fragment of human β-globin gene contains the last 20 bases of exon 2, all of intron 2 and exon 3, and the 3′ flanking region with the polyadenylate signal (27). The plasmid harboring the final transgene construct (paP2A2bGlo) was sequenced to confirm the integrity of the HMGA2 ORF. The 7.5 kb transgene fragment was released from the plasmid by HindIII digestion and purified by CsCl density centrifugation before microinjection. The transgene harboring the truncated HMGA2 ORF was made by a PCR-based methodology, using paP2A2bGlo as a template. The purified transgenes were microinjected into F₂ zygotes isolated from superovulated hybrid (C57BL/6J × CBA/J) F₁ females intercrossed with hybrid (C57BL/6J × CBA/J) F₁ males. The transgene was directly microinjected into the male pronuclei of the zygotes as described (28). The screening of the founder mice was done by Southern blotting using the transgene-specific probe sequence identified in Fig. 1A, and genomic PCR using transgene-specific primers as follows: forward 5′-TCATAGCACCCTCCTGTGCTGCA-3′ and reverse 5′-CATCAAGGGTCCCATAGACTCACCC-3′.

Mouse dissections and tissue processing. All procedures done on mice were according to the strict guidelines set forth by the Institutional Animal Care and Use Committee at University of Medicine and Dentistry of New Jersey. For histologic examination, the tissues were fixed in 4% paraformaldehyde. For RNA isolation, the tissues were flash-frozen and later processed with the RNEasy kit (Qiagen, Inc., Valencia, CA).

Transgene expression. The expression of the transgene was assessed by Northern blotting and/or RT-PCR analysis using transgene-specific primers: for full-length HMGA2 transgene, forward 5′-GTCTGCCGAAGAGGACTAGGC-3′ and reverse 5′-CACACAGACCAGCACGTTGC-3′; for the truncated HMGA2 transgene, forward 5′-CAAGAGGCAGACCTAGGAAATAGATC-3′ and reverse 5′-CACGTTGCCCAGGAGCC-3′. For Northern blot analysis, the probe (7) recognized an RNA species of mobility of 0.7 kb for the transgene and 4.1 kb for the endogenous HMGA2.

Immunohistochemistry. Paraffin-embedded samples were sectioned at 4 μm and stained for routine histology by H&E (Fisher Scientific, Pittsburgh, PA) and with polyclonal anti-HMGA2 antibody (7) using an immunoperoxidase technique (Vectastain ABC Elite kit, Vector Laboratories, Burlingame, CA) using the protocol from the supplier. Tissues from Hmga2-null mice (7) were used as negative controls.

Generation of transgenic mice. To direct the expression of the HMGA2 transgene to the differentiated adipocytes, a 330 bp cDNA fragment encoding full-length human HMGA2 was placed under the control of 5.4 kb aP2 promoter fragment (ref. 26; Fig. 1A). The transgenic founders were screened by Southern blot analysis. Fifteen founder mice were obtained, but only one transmitted the transgene to only 2 (of 31) of its progeny, which in turn failed to breed. Consequently, all phenotypic analysis was restricted to 11 founder mice and two first-generation transgenic offspring. The expression of the transgene was assessed by Northern blotting and confirmed by transgene sequence-specific RT-PCR. Because mesenchymal cells are present in most tissues and organs (29), transgene expression was observed in a number of tissues and, as expected, the endogenous HMGA2 was not detected in adult tissues (Fig. 1B; Table 1; ref. 5).

HMGA2-transgenic mice give rise to benign mesenchymal tumors. A subset of the transgene-expressing tissues exhibited a neoplastic phenotype (Table 1) and all of the neoplastic tissues showed a high level of transgene expression. It should be noted that the appearance of the neoplastic phenotype was highly selective, as some of the tissues did not show neoplasia despite high transgene expression (e.g., testis, heart, and spleen in Fig. 1B), which shows cellular specificity with regard to the ability of HMGA2 expression to bring about conversion to tumorigenic phenotype. Eight of 11 transgenic founders, and both the F₁ progenies, produced tumors of the breast (Table 1; Fig. 1C). Interestingly, these mice included seven of nine males as well as three of four females. These tumors were well circumscribed, and in 8 of 10 cases multiple tumors were present bilaterally. Histopathologically, the breast tumors showed marked proliferation of both stromal and epithelial components, and thus diagnosed as fibroadenomas (Fig. 2). Pericanalicular growth pattern with stromal homogeneity was seen in the fibroadenomas. The representative tumor shown in Fig. 2B is exhibiting a pericanalicular growth pattern, where both ductal and stromal components are abundant. The male founder breast fibroadenoma in Fig. 2D is showing a similar mixed growth pattern. Mitotic figures were not seen in any of the tumors, ruling out the diagnosis of phyllodes tumor.

Five of 11 founder mice and one of the F₁ progenies presented with bulging palpable tumors in the neck area in bilateral manner (Table 1). Upon dissection and histopathologic examination, these were diagnosed as salivary gland adenomas. All the salivary gland tumors tested showed expression of the transgene (an example is shown in Fig. 1B). All of the salivary gland tumors were originated from the serous type (parotid) glands (Fig. 3B). Although none of the mice produced adenomas of the sublingual and submandibular type glands, the mucinous component in the mixed-type submandibular gland of one mouse showed a hyperplastic phenotype. It is interesting to note that the majority of the human salivary gland adenomas also originate in the parotid glands (30). Furthermore, all of the adenomas showed similarities with the classic human pleomorphic adenomas of the salivary glands, with balanced proportion of both islands of epithelial ductal structures and interposing myxoid stroma. The adenoma from one mouse exhibited hamartomatous changes (data not shown).

Six of the male founders and one of the F₁ progenies showed hyperplasia and dilation of the ducts of the preputial gland (Fig. 3D). None of the females produced this phenotype. Preputial gland is an androgen-responsive modified sebaceous gland, and rodent preputial glands are used as models of human sebaceous glands (31). All the hyperplastic preputial glands in the founder mice showed high levels of transgene expression. Three of four females also exhibited overgrowth of the uteruses. Upon histologic examination, these tissues showed marked endometrial hyperplasia.

Transgene expression is restricted to stroma. The immunohistochemical analysis of HMGA2 expression in the neoplastic tissues of HMGA2-transgenic mice revealed that the transgene was expressed exclusively in the stromal component of the breast fibroadenomas, salivary gland adenomas, and preputial gland hyperplasias, whereas the epithelial components of all of these tumor types were negative for HMGA2 expression (Fig. 4). This observation further confirms that the aP2 promoter specifically targeted the transgene to mesenchymal cells. Additionally, the mesenchymal stromal cells also did not express E-cadherin (data not shown).

Truncated HMGA2 transgenic mice produce neoplastic phenotype. To test the effects of misexpression of truncated HMGA2, transgenic mice were produced by microinjecting the aP2-hHMGA2-hβGlobin transgene that was mutated so that the coding region of HMGA2 was truncated immediately after the third AT-hook DNA-binding domain (Fig. 1A). The rest of the transgene was exactly the same as the one harboring the full-length HMGA2 coding region.

The frequency of obtaining truncated HMGA2 transgenic mice among live births was the same as obtained with the full-length HMGA2 transgene microinjections. The positive founders included two females and one male. All of these mice also failed to breed and no transgenic line could be established. As for the full-length transgene, the truncated transgene also showed expression in a wide range of tissues (data not shown).

Two of the three founder mice expressing the truncated form of HMGA2 showed fibroadenomas of the breast (Fig. 5A), whereas the third one showed cystic ductal dilation in the breasts. One mouse showed salivary gland hyperplasia. Both of the females had endometrial hyperplasia (Fig. 5B-D), whereas the male mouse showed dilation of the ducts of preputial gland. Thus, the transgenic mice misexpressing the truncated form of HMGA2 produced similar phenotypic abnormalities as the mice misexpressing the full-length HMGA2.

Recurrent involvement of HMGA2 gene translocations in benign mesenchymal tumors has prompted speculation that HMGA2 gene might play a causal role in the pathobiology of a subset of these tumors (13, 32). However, in vivo evidence has been lacking that would directly link derepression of HMGA2 in adult mesenchyme to generation of majority of human solid mesenchymal tumor types found to carry HMGA2 gene rearrangements. Recently, two transgenic mouse models have been described that overexpress a truncated version of HMGA2, containing only the three AT-hook domains and are devoid of the acidic carboxy terminus, under the constitutive cytomegalovirus (CMV) and H-2K^b promoters (22, 33). Arlotta et al. (33) showed only lipomas in a small percentage of the transgenic mice, whereas no other mesenchymal tumors were observed in either model. Another transgenic mouse model expressed full-length mouse HMGA2 under the CMV promoter, but only exhibited pituitary adenomas of epithelial origin (34). It is important to note that all these transgenic models have HMGA2 expression under constitutive, rather than mesenchyme-specific, promoters. Thus, several questions remained unresolved regarding the causal role of HMGA2 misexpression in mesenchymal tumorigenesis.

The study presented here describes a transgenic mouse model that misexpresses HMGA2 specifically in the differentiated mesenchymal cells. As a result, the transgenic founders expressed the full-length human HMGA2 in a number of tissues that produced, in a subset of these tissues, a variety of neoplastic phenotypes; fibroadenomas of the breast, salivary gland adenomas, preputial (sebaceous) gland hyperplasia, and endometrial hyperplasia. Importantly, these tumor types faithfully mimic many of the tumor types that are observed in humans that have recurrent rearrangements involving HMGA2 gene, including the aforementioned fibroadenomas of the breast (9, 13, 32), pleomorphic adenomas of the salivary glands (14–16), and endometerial polyps (18, 19). Another strength of the fidelity of the mouse model is the selectivity of the neoplastic phenotype, as some of the tissues did not show neoplasia despite high transgene expression, for example, testis, heart, and spleen (Fig. 1B). Human tumor types arising in these tissues also have never been reported to show HMGA2 misexpression, which shows a cellular specificity with regard to the ability of HMGA2 misexpression to bring about conversion to tumorigenic phenotype in both species. However, pulmonary hamartomas were not observed in the mouse model and this could simply reflect species-specific differences between human and mouse tumorigenesis. Intriguingly, our mouse model exhibited preputial gland hyperplasia, which suggests that it would be of interest to investigate the involvement of HMGA2 misexpression in human sebaceous gland neoplasia.

Our HMGA2 transgenic mice not only replicated the tumor specificity found in humans as a result of misexpression of HMGA2, but the mouse tumor pathology was also remarkably similar to the human counterpart. All of the salivary gland adenomas were originated from the parotid glands, whereas none of the mice produced adenomas of the sublingual and submandibular type. This closely parallels the origin of the majority of the human salivary gland adenomas exhibiting a translocation at chromosome 12q13-15, which also originate in the parotid glands (30). Furthermore, all of the adenomas showed similarities with the classic human pleomorphic adenomas of the salivary glands, with balanced proportion of both islands of epithelial ductal structures and interposing myxoid stroma. These results provide in vivo evidence of causality of HMGA2 misexpression in the genesis of human benign mesenchymal neoplasia.

The immunohistochemical studies of the transgenic HMGA2 expression in the neoplastic tissues showed that expression was limited to the stromal component of the tumor and virtually absent in the epithelial component. Nevertheless, there was significant proliferation of the epithelial component of the biphasic tumors. This is similar to the observation made in human tumors harboring translocation at HMGA2 locus. Tallini et al. (35) have shown that the tumor types that are composed of both epithelial and mesenchymal components exhibit expression of the truncated HMGA2 only in the stromal cells.

Recently, the importance of the stromal microenvironment in the initiation and progression of epithelial cancers has started receiving deserved attention (36). It is suggested that paracrine factors produced by the stromal cells can induce epithelial transformation. Furthermore, specific genetic alterations in the stromal cells may give rise to premalignant and malignant epithelial tumors (37). Our results strongly support the idea that misexpression of HMGA2 in the stromal (mesenchymal) component of the breast and salivary gland is the causative factor for epithelial proliferation and the ultimate generation of solid tumors in these tissues. It can be postulated that the proliferation of the epithelial component in these tumors is a result of paracrine signals produced by the stromal cells that aberrantly express HMGA2. Thus, it can be envisaged that HMGA2 might be a critical component of mesenchyme-driven epithelial proliferation and differentiation. Further work would be required to elucidate this pathway.

It has been postulated that the truncation of HMGA2 and/or addition of ectopic sequences after the third AT-hook are necessary for its oncogenic ability (22, 23). Conflicting evidence to this view has arisen as a subset of the tumors that carried rearrangements at 12q13-15 did not disrupt the coding region of HMGA2 gene and expressed full-length HMGA2 protein (16, 24, 25). We provide the first direct in vivo evidence that misexpression of HMGA2 protein is sufficient for induction of breast fibroadenomas and salivary gland adenomas, which showed a histopathology that closely resembled that of their human counterparts. The neoplastic disposition of the HMGA2 transgenic mice strongly suggests that full-length HMGA2 has oncogenic potential and its truncation is not necessary for cellular transformation in benign mesenchymal tumors. Thus, it can be concluded that mere misexpression of HMGA2 in a differentiated mesenchymal cell is sufficient to bring about neoplastic change, regardless of the structural integrity of its acidic carboxy terminal tail or presence of add-on ectopic sequences. Instead, our results support the idea that translocation breaks disrupt regulatory sequences in the HMGA2 gene resulting in derepression, and thus misexpression, and stabilization of the resulting transcript (38). Cumulatively, the data from the present and previous studies suggest that the minimum requirement for benign mesenchymal tumorigenesis is the misexpression of the three DNA-binding domains of HMGA2.

Grant support: National Cancer Institute/NIH grant CA77929 (K.K. Chada).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.