The HMGA family consists of four members, each containing several ‘AT-hooks’, the functional motif characteristic of this protein family. Through these hooks, HMGAs bind preferentially to AT-rich stretches in B-form DNA and induce conformational changes that promote subsequent recruitment of additional components to the binding sites (Figure 1). HMGA proteins have an acidic C-terminal tail, which could be important for protein-protein interactions [1,2].

The HMGB family consists of three variants, each containing two functional motifs (HMG boxes) and a highly acidic C-terminal. The HMG boxes are formed by three α-helices folded together to an L-shaped structure part of which penetrates the minor groove and sharply bends the DNA. Slight differences between the HMG boxes confer specificity to the various HMGB proteins, whereas the acidic tails modulate their affinity for a variety of distorted DNA structures (Figure 1).

The HMGN protein family is characterized by a positively charged domain, the nucleosome binding domain, and by an acidic C-terminal, the chromatin unfolding domain. HMGNs bind specifically to nucleosomes and alter both the local and the higher order structure of the chromatin fiber.

HMG proteins bind to chromatin without any known preference for the underlying DNA sequence; their functional specificity could depend on interactions with specific regulatory factors, or on their ability to continuously target a specific subset of chromatin conformations. Possible mechanisms by which HMGs alter the architecture and function of chromatin are depicted in Figure 1. Several reviews provide extensive information on the structure and architectural functions of HMG proteins [1-7].

On the basis of many in vitro experiments, HMGs are viewed as architectural components that remodel the structure of chromatin. More recent imaging experiments with living cells show that the interaction of these proteins with chromatin is highly dynamic, that they are not stably associated with specific sites and that each HMG molecule can sample the nucleosomes for potential binding sites. In living cells, HMG molecules continuously hop from one binding site to another in a ‘stop and go’ fashion. The average ‘stop’ stage lasts several seconds and is longer than the ‘go’ stage; therefore, most of the time the HMG molecules are bound to chromatin. However, at the level of local binding sites there is a continuous turnover of HMG proteins. Furthermore, each HMG protein competes for chromatin binding sites with other members of the same family and with histone H1. Thus, in living cells HMGs remodel chromatin as members of a dynamic network of chromatin-binding proteins, which are likely to include other proteins transiently binding to chromatin. The kinetic properties of chromatin-binding proteins such as the HMGs provide a mechanism for functional redundancy among related proteins. The function of a missing protein can be partially compensated by a closely related protein, as is the case for histone H1: deletion of one variant leads to increased synthesis of other histone H1 variants.

HMGA overexpression is a hallmark of several malignant and benign tumors [1,4] (Table 2). The faulty expression of many genes in malignant and benign human tumors is a direct result of HMGA overexpression [4,15,16]. In human epithelial cells, HMGA1 promotes tumor progression and epithelial-mesenchymal transition (EMT) [17]; EMT occurs during movements of embryonic cell layers and during metastasis. Likewise, recent studies indicate that HMGA2 is an important mediator of transforming growth factor (TGF-β) signaling that controls epithelial differentiation, tumor invasiveness and metastasis [18].

HMGA1 and HMGA2 are both implicated in the etiology of lipomas (benign tumors composed of mature fat cells) [1]. Rearrangements of the HMGA2 gene involving 12q13-15 chromosomal translocations are perhaps the most common rearrangement in human benign tumors of mesenchymal origin. Many studies link these translocations and overexpression of HMGA2 to several types of benign tumors. Overexpression of HMGA2 is sufficient to induce benign tumors in mice [19].

The involvement of HMGAs in specific gene expression is crucial for their diverse roles in differentiation and tumorigenesis. In humans, HMGA overexpression in malignant and benign tumors causes aberrant expression of genes that regulate cell proliferation or apoptosis [4,15,16]. Recently described interacting partners, such as nuclear factor (NF)-κB [20] and the tumor suppressors p53 [21,22] or pRB [23], provide a clue to how HMGA proteins could influence gene expression in cancer. However, HMGA proteins also inhibit nucleotide excision repair [7,24], which could mean that HMGA proteins promote tumorigenesis by increasing genome instability.

Besides their obvious roles in oncogenesis, an antioncogenic potential was recently attributed to HMGA1 proteins. Antiproliferative activities were observed in lymphoproliferative disorders [9] and during cellular senescence [25], which prevent rather than promote malignant transformation. In the context of cellular senescence, HMGAs modulate global chromatin architecture by contributing to the formation of senescence-associated heterochromatin foci.

In mice, there are three HMGB variants, HMGB1, HMGB2 and HMGB3, all of which are expressed in early embryos. During embryonic development, HMGB2 and HMGB3 are selectively down-regulated: HMGB2 remains highly expressed in lymphoid organs and testes, whereas HMGB3 is expressed mainly in primitive hematopoietic cells [28-30]. By contrast, HMGB1 continues to be expressed in all embryonic and adult cells [31]. The biological significance of the down-regulation of the HMGB variants is not fully understood. Induced overexpression of HMGBs alters the cellular transcription profile through interaction with transcription factors; however, no systematic studies on the effects have been done [5].

Studies with Xenopus eggs and Drosophila embryos, which contain large amounts of maternal HMGB but little or no histone H1, provide insights into one possible developmental function of these proteins. During early development of these organisms, the HMGB1:H1 ratio switches and increased histone H1 levels lead to changes in chromatin structure and gene expression [32,33]. These observations suggest that HMGB1 has a role in maintaining the chromatin structure, especially in early embryogenesis. However, neither HMGB1 protein nor mRNA is stored in the mouse oocyte, suggesting that the role of HMGB1 in the early embryogenesis of mammalians is different from those in other species [34].

Analyses of animals with altered HMGB levels reveal functional specificity among the various HMGB variants (Table 1). Thus, although all HMGB variants are structurally similar and induce similar architectural changes in chromatin, they have specific biological functions. The molecular mechanisms conferring functional specificity to the HMGB variants have not been studied in detail. They might reflect variant-specific interactions with regulatory molecules that target the HMGB variants to specific chromatin sites.

Like HMGA proteins, HMGBs are involved in regulating DNA repair processes [7]. However, HMGB proteins either promote lesion repair or induce apoptosis. Moreover, HMGBs preferentially bind to cisplatin-modified DNA or to misincorporated nucleoside analogs and consequently inhibit nucleotide excision repair by steric hindrance, a fact that is of great value in cancer treatment [35,36]. Hmgb1^-/- mice have increased chromosomal instabilities [6], an observation that links aberrant expression levels of HMGB proteins to the chromosomal instabilities frequently found in cancer cells.

Like HMGBs and HMGAs, the HMGN1 and HMGN2 genes are ubiquitously and highly expressed in all embryonic tissues. During mouse embryogenesis, these two HMGN genes are progressively down-regulated throughout the entire embryo, except in committed but continuously renewing cell types undergoing active differentiation, such as the basal layer of the epithelium or in kidney cells undergoing mesenchyme to epithelium transition [43,44]. Likewise, experiments with tissue culture cells show that HMGN expression is down-regulated during myogenesis, erythropoiesis and osteogenesis [42,43]. In C2C12 myoblast cells, aberrant expression of Hmgn1 inhibits myotube formation [45], whereas in primary limb bud mesenchymal cells overexpression of HMGN1 inhibits chondrocyte differentiation [43]. Depletion of HMGNs in one-cell or two-cell mouse embryos delays subsequent embryonic cleavages [46].

In Xenopus, HMGN proteins are absent from cleavage stages and synthesis starts with the activation of the embryonic genome at the midblastula transition (MBT). Alteration of HMGN levels after, but not before, the MBT leads to altered expression of mesoderm-specific genes and to developmental defects [42]. Taken together, the available data indicate that proper differentiation requires regulated levels of HMGN expression. It is not yet clear whether HMGNs affect the expression of a subset of specific genes or whether they act as general cofactors that optimize cell-specific transcription profiles.

So far only Hmgn1^-/- mice have been generated and they appear normal [47]. However, they are subfertile and hypersensitive to various stress conditions, such as exposure to UV light or ionizing irradiation [47-49] and show abnormalities in the development and maintenance of the corneal epithelium [49]. These abnormalities could be linked to the high expression of HMGN1 in the basal layer of the corneal epithelium where it colocalizes with p63, a protein involved in the regulation of epithelial differentiation [49]. Cells derived from Hmgn1^-/- mice have an altered transcription profile and are hypersensitive to stress [47,50]. It is likely that partial functional redundancy with HMGN2 dampens the deleterious effects of loss of HMGN1 [43].

The incidence of tumors in Hmgn1^-/- mice is almost twice that of wild-type mice, and Hmgn1^-/- cells have an increased tumorigenic potential, as measured by colony formation in soft agar and generation of tumors in nude mice [48]. This increased tumorigenic potential could be due to faulty DNA repair. Hmgn1^-/- mice and cells are hypersensitive to UV and ionizing radiation [47,48]. The UV hypersensitivity of Hmgn1^-/- cells is related to the ability of HMGN to ‘unfold’ chromatin, leading to the suggestion that HMGN increases the accessibility of the damaged DNA sites to the repair machinery [47]. HMGN1 could also have a unique role in transcription-coupled repair because it is specifically recruited by Cockayne syndrome protein A to the RNA polymerase stalled at the UV-damaged sites [51]. The hypersensitivity of Hmgn1^-/- mice to ionizing radiation might be due to their impaired ability to activate the G2-M checkpoint [48], a fault leading to increased genomic instability and cancer. HMGNs might influence DNA repair processes by altering the ability of repair factors to reach the damaged sites or by affecting chromatin modifications necessary for proper repair. Thus, given that HMGN1 modulates the rate of stress-induced histone modifications, it is possible that it also affects the modification of the core histones during DNA repair.

The mouse Hmgn1 gene is located on chromosome 16 in a region syntenic with the human Down syndrome region on chromosome 21 [52]. Cells from Down syndrome patients and from the Ts1Cje mouse, which serves as an animal model for this disease, carry three alleles of the HMGN1 gene and expresses elevated levels of HMGN1 protein [53]. The exact role of HMGN1 in Down syndrome is not understood; however, given the complex phenotype of this syndrome, which includes faulty DNA repair, and given the pleiotropic effects on the chromatin caused by altered HMGN1 expression, it is possible that elevated HMGN1 is a contributing factor in the development of this prevalent birth defect.

A common property of all HMGs is that during embryonic development their expression is gradually down-regulated (Figure 2). This down-regulation could be related to changes in chromatin organization and in the cellular transcription potential that occur during development. In several experimental systems, up- or down-regulation of HMG expression results in significant changes in the developmental program, suggesting that proper development and differentiation requires regulated levels of HMG proteins. However, so far, analysis of HMG knockout mice indicates that loss of a single HMG protein does not significantly impair embryonic development, because mice lacking only one type of HMG protein are born alive. Functional redundancy among members of the same HMG family probably compensates for loss of each HMG variant, thereby enabling proper embryonic development. It will be of interest to examine the phenotype of compound mutant mice lacking more than one HMG variant.

Once born, all the HMG knockout mice develop distinct, and in some cases very severe, phenotypes (Table 1), including increased levels of various tumors (Table 2). It is conceivable that cellular mechanisms involving HMG functional redundancy are not as efficient in fully differentiated tissues, where the levels of these proteins are relatively low, as in developing tissues. The phenotypes of the HMG knockout mice suggest functional specificity of HMG variants in the differentiated tissues. Several mechanisms could account for functional specificity of HMG variants. One possibility is tissue-specific HMG variant expression. For example, HMGB2 is expressed in adult testis and Hmgb2^-/- mice have defective spermatogenesis [29]. Alternatively, the chromatin structure of tissue-specific genes could be selectively susceptible to the architectural effects induced by the various HMG variants. However, the most likely possibility is that HMGs interact with tissue-specific regulatory factors that target the proteins to specific sites. Indeed, several reports indicate that all HMGs interact with a wide range of nuclear proteins. Thus, the biological effects of HMGs might depend on the availability of their interacting partners, which could vary among cells and which is not necessary constant throughout development. Variability in the levels of the HMG-interacting partners could explain the large range of biological functions affected by HMGs.

Furthermore, the main phenotypes observed in the HMG knockout mice could mask additional developmental defects. For example, it is likely that the early death of the Hmgb1^-/- mice as a result of hypoglycemia [5] masks additional abnormalities in these mice. Hmgn1^-/- mice that do survive show a wide range of phenotypes, including defects in the repair of DNA damaged by UV or ionizing radiation, defective stress responses and abnormal development of the corneal epithelium [47-50].