High mobility group A2 is a target for miRNA-98 in head and neck squamous cell carcinoma

High mobility group A2 (HMGA2) protein is a non-histone architectural transcription factor, which is a member of the HMGA family. This family is constituted by HMGA1a, HMGA1b, HMGA1c, and HMGA2[1, 2]. HMGAs are chromosomal proteins that bind through their AT binding motifs or AT hooks to the minor groove of AT-rich DNA strands[3]. The two types of HMGA protein, HMGA1 (HMG-I/Y) and HMGA2 (HMGI-C), have similar functions and are coded for by two different genes at chromosomal loci 6p21.3 and 12q15, respectively. HMGA proteins are expressed predominantly during embryogenesis, their expression being absent or greatly reduced in adult tissues [4, 5]. However, HMGA proteins have been implicated in the regulation of transcription, differentiation, neoplastic transformation, and integration and expression of viral genomes [6, 7]. HMGA2 has been implicated mainly in benign and malignant mesenchymal tumors[8] Nevertheless, increased expression of HMGA2 in oral squamous cell carcinoma has been linked to epithelial-mesenchymal transition, whereby epithelial cells acquire mesenchymal characteristics [8]. Additionally, HMGA2 expression in patients with breast cancer has been shown to be associated with poor prognosis and metastasis [9]. However, the precise role of and the molecular events elicited by HMGA2 in tumorigenesis still need to be defined. Furthermore, HMGA2 expression has been shown to be associated with enhanced selective chemosensitivity towards the topoisomerase (topo) II inhibitor, doxorubicin, in breast cancer HS578T and salivary Pa-4/HMGA2 cells [2]. These studies have suggested that HMGA2 induces a persistent basal H2AX Ser139 phosphorylation and perturbs doxorubicin-elicited DNA damage checkpoint promoting enhanced chemosensitivity towards doxorubicin treatment in HMGA2-expressing cells.

Although proteins have been established as regulatory factors in growth, development and accommodation to alterations in environment, our knowledge of the contribution of regulatory small RNA molecules to such is rapidly evolving [10]. MicroRNAs (miRNAs) are 20–24 nucleotide non-coding RNA molecules that post-transcriptionally control gene expression. They are spawned by RNA polymerase II as large primary miRNA (pri-miRNA) transcripts [11, 12]. Some miRNAs are structured in clusters and transcribed as polycistrons [13], while others exist in introns of protein-coding genes [14] and are co-transcribed with host genes. Pri-miRNA transcripts form typical fold-back structures that are identified and processed into ~70nt imperfect stem-loop miRNA precursors (pre-miRNAs) by a microprocessor complex [15, 16]. After processing by the microprocessor complex, pre-miRNAs are exported by Ran-GTP dependent transporter exportin-5 [17‐19] from the nucleus into cytoplasm, where they are processed by the RNAse III enzyme, Dicer, into imperfect dsRNA duplexes that contain both mature and complementary miRNA strands [20‐22]. One of the mature miRNA strands forms an effector ribonucleoprotein complex termed RISC (RNA induced silencing complex) which guides the miRNAs to specific mRNAs [23]. RISC either cleaves or blocks translation of the target mRNA, depending on the degree of sequence complementarity.

Here we report that HMGA2 expression in head and neck squamous cell carcinoma (HNSCC) cells is regulated in part by the expression of miRNA-98 (miR-98). Albeit HMGA2 is associated with enhanced selective chemosensitivity towards topoisomerase (topo) II inhibitor, doxorubicin in HNSCC, the expression of HMGA2 is thwarted by hypoxia. This is accompanied by enhanced expression of miRNA-98 and other miRNAs, which predictably target HMGA2. Moreover, we show that transfection of pre-miR-98™ during normoxia diminishes HMGA2 and potentiates resistance to doxorubicin and cisplatin. These findings implicate the role of a miRNA as a key element in modulating tumors in variable microenvironments.

Poor oxygenation (hypoxia) is present in the majority of human tumors and is associated with poor prognosis due to the protection it affords to radiotherapy and chemotherapy [25]. Here, we sought first to verify the chemoresistance effect of hypoxia to doxorubicin and cisplatin in SCC cell lines, comparing cell lines grown in normoxic and hypoxic atmospheres. These studies revealed all SCC cell lines studied here using MTT cell vitality responded in a dose-dependent manner to doxorubicin. Moreover, when the cell lines were grown under hypoxic conditions, they demonstrated enhanced cell viability (Figure 1). Similar results were also observed for cisplatin (not shown). The mechanisms for chemoresistance to these agents under hypoxia were initially unclear. Existing knowledge dictates that tumors cope with hypoxic microenvironments by bringing forth multiple cellular response pathways that alter gene expression and affect tumor progression by strongly suppressing the rates of mRNA translation during hypoxia [25].

Although the roles of signaling cascades and proteins as regulatory factors in development, neoplasia and adaptation to the environment are becoming well established, the evidence for involvement of regulatory small RNA molecules as important regulators of both transcriptional and posttranscriptional gene silencing is presently mounting [10]. Thus, in the present study we grew SCC cell lines both in a normoxic atmosphere and in an environment that mimics the hypoxia encountered in solid head and neck tumors and used differential display hybridization to a human miRNA chip, and Real-time RT-PCR and Northern blot validation, to discern miRNA expression profiles. SCC cells that were grown under hypoxic conditions differentially expressed 47 miRNA genes compared to cells grown under normoxic conditions (P < 0.04). The following miRNAs were expressed to the greatest extent during hypoxia:hsa-miRNA-572, -214, -563, -637, -628, -191, -210, -498, -373, -98, -148b, -15a, -148-a, -200a, -30b, -429, -7, hsa-let-7e, -7i, and -7g (Table 1). Conversely, the following miRNAs were most diminished in their expression in hypoxic cells: hsa-miRNA-122a, -565, -195, -30e-5p, -374, -19a, -101, -424, -29b, -186, -141, -320, -422b, and -197 (Table 1).

One of the most prevalent changes in expression profiles involved a 4.56 fold increased expression of hsa-miR-98. To validate these findings we grew SCC cell lines in either a normoxic or hypoxic environment as before and assessed the miRNA-98 levels with miRVana qRT-PCR miRNA™ Detection kit. These studies revealed that following hypoxia the relative levels of miRNA-98 increased among the SCC cell lines in an oxygen dependent manner (Figure 2). These changes in miRNA-98 were validated in Northern blot analyses using mirVana miR-98™ Probe, normalized to 5S rRNA (not shown).

To further appraise the role of miRNA-98 during normoxia and hypoxia SCC cell lines were transfected with the pre-miR-98™ or anti-miR-98™ (Ambion) using siPORT™ NeoFX™ Transfection Agent, for 72 hours. The predicted changes in the levels of miRNA-98 were validated by miRVana qRT-PCR™ as before (not shown). SCC cell lines transfected with either anti-miR-98™ or with pre-miR-98™ were then challenged with varying doses of doxorubicin and the surviving fraction of cells as a percentage of control was determined using the MTT assay. Cell lines grown in normoxic conditions revealed greater percent of surviving cells when treated with pre-miR-98™ compared to controls and anti-miR-98™ transfected cell lines. There was no significant difference between cell lines treated with anti-miR-98™ and controls. Cell lines grown under hypoxic conditions and transfected with pre-miR-98™ showed a slight increase in surviving cells compared to control cells and anti-miR-98™ transfected cell lines. There was no significant difference between control and anti-miR-98™ transfected cell lines (Figure 3a). Similar findings were also extended to other double-strand breaks elicited by cisplatin in SCC cells (Figure 3b). These studies revealed that normoxic cell lines treated with pre-miR-98™ acquired chemoresistance to doxorubicin and cisplatin.

Next, we sought to assess potential targets for miRNA-98 that might explain the observed chemoresistance to these drugs during hypoxia or in pre-miR-98™ transfected normoxic cell lines. In that miRNA-98 has 487 conserved miRNA:UTR pairs within the human genome by miRNA viewer, we selected potential targets that were predicted from at least two of the database resources with at least two miRNA target sites. Utilizing miRNA viewer, the high mobility group protein HMGI-C (High mobility group AT-hook 2, HMGA2) was one of the highest ranking genes with five potential target sites for hsa-miRNA-98. Utilizing other databases it was found that PicTar defined 6 hsa-miRNA-98 target sites, and TargetScanS identified 7 hsa-miRNA-98 target sites (Figure 4). Notable was that a number of other miRNAs included in the hypoxia profile also were potential regulators for HMGA2. Most prevalent among this later group utilizing TargetScanS were hsa-let-7g which possessed 7 target sites for HMGA2 (Figure 5), hsa-let-7e with 7 target sites (Figure 6) and hsa-let-7i with 7 sites (Figure 7). Albeit, the expression of has-let-7g, -7i, and -7i was similar to miRNA-98, being low in normoxia, but increasing following hypoxia (Table 1).

Recognizing that HMGA2 has been shown to perturb doxorubicin-elicited DNA damage checkpoint control and to promote enhanced chemosensitivity towards doxorubicin and cisplatin treatment in HMGA2-expressing breast cancer cells [2], we assessed whether hypoxia altered the expression of HMGA2; SCC cell lines were subjected to hypoxia, harvested and the expression of HMGA2 was determined by Western blot. These studies revealed that hypoxia diminished the levels of HMGA2 protein expression (Figure 8a).

To verify that HMGA2 potentiated genotoxic stress in SCC cell lines we paralleled our studies to those performed in breast cancer cells [2]. In essence, cell lines were transfected with siRNA oligonucleotides for HGMA2 as described above. The oligonucleotides encoding HMGA2 siRNA was compared with an oligonucleotide scramble siRNA as control [2]. These studies revealed that HMGA2 siRNA induced down-regulation of HMGA2 protein expression as assessed by Western blot (Figure 8b lower panel), and diminished doxorubicin sensitivity by MTT assays (Figure 8c). Recognizing that SCC cell lines became resistant to doxorubicin during periods of hypoxia and following diminished HMGA2 expression after siRNA transfection in normoxic cells we sought to verify that HMGA2 was a feasible target for miRNA-98 in SCC cell lines. To further appraise HMGA2 as a miRNA-98 target, SCC cell lines were transfected with the pre-miR-98™ (Ambion) using siPORT™ NeoFX™ Transfection Agent, for 72 hours. The increased levels of miRNA-98 were validated by miRVana qRT-PCR™ as before. These treatments resulted in diminishment in protein levels and mRNA expression of HMGA2 in control the normoxic SCC-4 cell line as well as cell lines exposed to hypoxia (Figure 9a,b); similar results were obtained for SCC-10B and SCC-9 cell lines (not shown). The demonstration that pre-miRNA-98™ diminished mRNA expression of HMGA2 was surprising, since the tenet of early studies, using reporter constructs had indicated that complete complementarities between miRNAs and target mRNAs was necessary for cleavage, whereas mismatches result in translational repression. However, the findings reported here are supported by more recent research, which revealed that miRNAs with partial complementarities can induce attenuation in cognate transcript levels [26‐29]. In fact, most recently it has been shown that the ability of miRNAs to expedite poly(A) removal does not result from decreased translation; nor does translational repression by miRNAs require a poly(A) tail, with a 3' histone stem-loop being an effective substitute [30].

Conversely, the use of the anti-miR-98™ provided little information in the present studies. Cell lines grown under normoxia express little miRNA-98 so that the effects of anti-mir-98™ on HMGA2 levels were limited (Figure 9a,b). Transfection with anti-miR-98™ had only minor effects on the HMGA2 levels in SCC cell lines grown in hypoxia (Figure 9a,b). This is probably owing to the increased expression during hypoxia of other targeting miRNAs such as hsa – let-7g, hsa-let-7e and hsa-let-7i all of which possess 7 complementary target sites for HMGA2. Inherent in these findings is the recognition that derepression of a mRNA targeted by a miRNAs in an altered microenvironment probably can not be achieved through down-regulation of a single miRNA in that most mRNAs possess multiple complementary target sites with a number of miRNAs.

Until lately, there was only speculation that inhibition by a miRNA to a specific mRNA could be effectively reversed. However, recent studies have shown that CAT-1 mRNA and reporters bearing its 3'UTR can be relieved from miR-122-mediated repression in human Huh7 hepatoma cells when they are subjected to different types of stress. Moreover, that the derepression is accompanied by the release of CAT-1 mRNA from cytoplasmic processing bodies and its entry into polysomes and that the process involves binding of HuR, trans-acting AU rich element (ARE) binding protein, to the 3'UTR of CAT-1 mRNA [31]. There is also additional evidence indicating that some miRNA-controlled mRNAs can be relieved from repression by synaptic stimulation [32, 33] indicating that miRNA regulation is more pronounced than previously anticipated and that it is able to respond quickly to specific cellular needs. What is yet to be determined is whether changes in oxygen tension provide a like mechanism for derepression of miRNAs against HMGA2. There is clear evidence that HuR, a major ARE-binding protein is expressed in many different cells and regulates the stability, subcellular shuttling, and/or translation of ARE-mRNAs (reviewed [34]). Moreover, the level of cytoplasmic HuR increases during cell proliferation, differentiation, and in response to stress stimuli [35‐38] decreasing with senescence [39]. Also, the high level of cytoplasmic HuR in cancer cells is in keeping with the ability of HuR to increase the stability and translation of ARE-mRNAs that regulate cell proliferation and neovascularization by stabilizing VEGF following hypoxia. HuR has also been shown to shuttle between nucleus and cytoplasm during hypoxia and to stabilize HIF-1α and EGFR mRNAs [34]. Consequently, we searched the AERD database [40] but were not able to demonstrate any of the five clusters containing adenylate uridiylate (AU)-rich elements in their 3'-untranslated region (3'-UTR) that would afford HMGA2 stabilization or derepression by HuR. Furthermore, since HuR shuttles from the nucleus to the cytoplasm during hypoxia it would appear that HMGA2 would be overwhelmed by HuR, making it unlikely that HMGA mRNA is stabilized like other ARE-mRNAs.

These studies validate the observation that HMGA2 plays a prominent role in governing genotoxic responses. However, this may only represent cells growing under normal oxygen tensions. The demonstration that miRNA profiles are altered during hypoxia and repress a genotoxic response indicates that changes in microenvironment in eukaryotes mimic lower species and plants, where, for example, abiotic stresses regulate the expression of thousands of genes in plants at both transcriptional and posttranscriptional levels through a number of miRNAs and other small regulatory RNAs[41]. Yet to be established is to what extent miRNA expression is regulated by defined signaling cascades and transcription factors.