A novel thyroid hormone receptor isoform, TRβ2-46, promotes SKP2 expression and retinoblastoma cell proliferation

Retinoblastoma is a childhood retinal tumor that develops from cone photoreceptor precursors in response to inactivating RB1 mutations and loss of functional RB protein. The cone precursor's response to RB loss involves cell type–specific signaling circuitry that helps to drive tumorigenesis. One component of the cone precursor circuitry, the thyroid hormone receptor β2 (TRβ2), enables the aberrant proliferation of diverse RB-deficient cells in part by opposing the down-regulation of S-phase kinase-associated protein 2 (SKP2) by the more widely expressed and tumor-suppressive TRβ1. However, it is unclear how TRβ2 opposes TRβ1 to enable SKP2 expression and cell proliferation. Here, we show that in human retinoblastoma cells TRβ2 mRNA encodes two TRβ2 protein isoforms: a predominantly cytoplasmic 54-kDa protein (TRβ2-54) corresponding to the well-characterized full-length murine Trβ2 and an N-terminally truncated and exclusively cytoplasmic 46-kDa protein (TRβ2-46) that starts at Met-79. Whereas TRβ2 knockdown decreased SKP2 expression and impaired retinoblastoma cell cycle progression, re-expression of TRβ2-46 but not TRβ2-54 stabilized SKP2 and restored proliferation to an extent similar to that of ectopic SKP2 restoration. We conclude that TRβ2-46 is an oncogenic thyroid hormone receptor isoform that promotes SKP2 expression and SKP2-dependent retinoblastoma cell proliferation.

Retinoblastoma is a childhood retinal tumor that develops from cone photoreceptor precursors in response to inactivating RB1 mutations and loss of functional RB protein. The cone precursor's response to RB loss involves cell type-specific signaling circuitry that helps to drive tumorigenesis. One component of the cone precursor circuitry, the thyroid hormone receptor ␤2 (TR␤2), enables the aberrant proliferation of diverse RB-deficient cells in part by opposing the down-regulation of S-phase kinase-associated protein 2 (SKP2) by the more widely expressed and tumor-suppressive TR␤1. However, it is unclear how TR␤2 opposes TR␤1 to enable SKP2 expression and cell proliferation. Here, we show that in human retinoblastoma cells TR␤2 mRNA encodes two TR␤2 protein isoforms: a predominantly cytoplasmic 54-kDa protein (TR␤2-54) corresponding to the well-characterized full-length murine Tr␤2 and an N-terminally truncated and exclusively cytoplasmic 46-kDa protein (TR␤2-46) that starts at Met-79. Whereas TR␤2 knockdown decreased SKP2 expression and impaired retinoblastoma cell cycle progression, re-expression of TR␤2-46 but not TR␤2-54 stabilized SKP2 and restored proliferation to an extent similar to that of ectopic SKP2 restoration. We conclude that TR␤2-46 is an oncogenic thyroid hormone receptor isoform that promotes SKP2 expression and SKP2-dependent retinoblastoma cell proliferation.
Cancers are caused by abnormalities in oncogenes or tumorsuppressor genes that initiate and advance tumorigenesis. At the initiation step, cell type-specific circuitry may sensitize cells to the initial oncogenic insult. Understanding how cell type-specific circuitry sensitizes to oncogenic changes may enable rational cancer prevention and treatment approaches.
Retinoblastoma is a childhood retinal tumor that has provided insights into the role of cell type-specific circuitry in tumor initiation (1). Most retinoblastomas are thought to arise from cone photoreceptor precursors in response to biallelic inactivation of the RB1 gene and loss of functional RB protein (2,3). Human cone precursor circuitry may sensitize to RB1 mutation via intrinsic high expression of oncoproteins, such as MYCN and MDM2, and cone lineage transcription factors, such as retinoid X receptor-␥ (RXR␥) 3 and thyroid hormone receptor ␤2 (TR␤2) (2,4). RXR␥ and TR␤2 normally mediate cone photoreceptor differentiation (5, 6) but promote cone precursor proliferation and retinoblastoma genesis after RB loss (2,4). RXR␥ enables retinoblastoma cell survival in part by inducing MDM2 expression via a human-specific MDM2 promoter element (4). However, the oncogenic role of TR␤2 is enigmatic.
TR␤2 is highly expressed in a limited number of cell types, including cone photoreceptor precursors, pituicytes, and cochlear hair cells (7)(8)(9), each of which aberrantly proliferates in response to RB loss (2,10,11). Indeed, TR␤2 is required for proliferation of retinoblastoma cells and enhances growth of Rb1-null mouse pituitary tumors, whereas ectopic TR␤2 enabled proliferation of RB-depleted neuroblastoma cells (12). TR␤2 appears to promote RB-deficient human retinoblastoma as well as Rb-deficient mouse pituitary tumors by antagonizing the highly related, more widely expressed, and tumor-suppressive thyroid hormone receptor TR␤1 (12,13). TR␤2 and TR␤1 are produced from the same THRB gene but use alternative transcriptional promoters and N-terminal coding exons (8,14). They both have canonical nuclear hormone receptor structure with an N-terminal "A/B" corepressor and coactivator-binding domain, a central DNA-binding domain, and a C-terminal T 3 -binding domain (Fig. 1A). However, their distinct A/B domains mediate distinct interactions and effects (15)(16)(17), including enhanced TR␤2-mediated transactivation (18). Additionally, TR␤1 can inhibit PI 3-kinase signaling (19,20) and was found to suppress liver and mammary tumors via induction of the NCoR transcriptional corepressor (21). Meanwhile, TR␤2 was found to oppose a TR␤1-dependent downregulation of the F-box protein SKP2 (12), which is required for production of RB-deficient tumors (22,23). However, it is unclear how TR␤2 opposes TR␤1 to enhance SKP2 expression and cell proliferation. Here, we report that these functions are mediated by a novel N-terminally truncated and cytoplasmlocalized TR␤2 isoform.
Because retinoblastomas are derived from cone precursors (2, 3) and TR␤2 is solely detected in cone precursors in the developing human retina (4, 24), we examined whether TR␤2-54 and TR␤2-46 are expressed in the cone precursor cell of origin. In Western blots, the main TR␤2 species in developing retina comigrated with retinoblastoma cell TR␤2-46, whereas a less abundant species comigrated with TR␤2-54 (Fig.  1B). Because of unavoidable sample limitations, a lower amount of fetal retina protein was loaded, and all bands, including the GAPDH loading control, migrated more slowly than their counterparts in retinoblastoma samples. A similar ratio of TR␤2-46 and TR␤2-54 comigrating species was detected in three retinae ( Fig. 1B and data not shown). The high TR␤2-46 and low TR␤2-54 in human retina differ from what was seen in mouse retina where only one specifically recognized species was reported (25). As cones comprise ϳ2-3% of human retinal cells, TR␤2-46 and TR␤2-54 are more highly expressed in cone precursors than appears from analyses of whole-retina lysates.   TR␤1 is mainly detected in the nucleus but can shuttle between cytoplasmic and nuclear compartments (26,27) and undergo T 3 -induced cytoplasm-to-nucleus translocation (20). TR␤2 is also thought to be mainly nuclear (26); however, by immunostaining, TR␤2 was perinuclear or cytoplasmic in later stages of mouse cone differentiation (25) and was mainly cytoplasmic in human cone precursors and retinoblastoma cells (2,4,24). To define the subcellular localization of the different TR␤2 isoforms, retinoblastoma cells were subjected to cytoplasmic and nuclear fractionation and TR␤2 immunoblotting. Separation of nuclear and cytoplasmic components was confirmed by detection of GAPDH solely in cytoplasmic fractions and Lamin B in nuclear fractions (Fig. 1D). As in past immunostaining analyses, the vast majority of TR␤2-46 and TR␤2-54 were in the cytosol in three retinoblastoma cell lines (Fig. 1D). However, after long exposures, TR␤2-54 and the nonspecific * species were also detected in nuclear fractions, whereas the more rapidly migrating TR␤2-46 was detected solely in cytoplasmic fractions (Fig. 1D).

TR␤2-46 translation initiates at methionine 79
The full-length human TR␤2 transcript corresponding to the well-characterized murine Tr␤2 is represented by GENCODE transcript ENST00000280696.9. This RNA is predicted to encode a polypeptide of 54.4 kDa (UniProt P10828, isoform ␤2). We previously found that transduction of RB177 cells with TR␤2 cDNA containing the same open reading frame mainly increased expression of TR␤2-54, based on its comigration with the major endogenous TR␤2 species (12). Thus, we sought to define the origin of the smaller TR␤2-46.
We first assessed whether TR␤2-46 resulted from differential splicing. To do so, we amplified cDNA from two retinoblastoma cell lines with forward primer F1 positioned at the 5Ј end of the predicted TR␤2 coding sequence and reverse primers R1-R6 in each downstream exon ( Fig. 2A). In both lines, each primer pair amplified a single PCR product of the predicted sizes ( Fig. 2B), suggesting that there were no novel splice sites between the known TR␤2 exons. We next evaluated whether alternative 5Ј exons are spliced to the TR␤2 exon by amplifying RB176 cDNA using reverse primer R1 and forward primers F2-F6 ( Fig. 2A). This generated PCR products of the predicted sizes using the F2 and F4 primers with 5Ј ends at nucleotides Ϫ102 and Ϫ301, respectively (Fig. 2, B and C). The 825-nucleotide PCR product made with the F4-R1 primers indicated that THRB RNA that encodes TR␤2 had a 5Ј-UTR of Ն301 nucleotides. This is 68 nucleotides longer than the murine ortholog (RefSeq NM_009380.3) (Fig. 2C) but within a previously deduced 377-nucleotide 5Ј-UTR in the mouse pituitary Tr␤2 transcript (14). The 825-nucleotide PCR product obtained using the F4-R1 primer pair was sequenced and confirmed to contain the predicted 301-nucleotide 5Ј-UTR (Fig. 2C). Thus, we confirmed that human retina expresses a THRB RNA encoding TR␤2 but did not detect novel splicing events that might produce TR␤2-46.

TR␤2-46 but not TR␤2-54 promotes SKP2-mediated cell cycle progression
Having identified the two TR␤2 isoforms, we examined their roles in retinoblastoma cell proliferation. In past analyses, TR␤2 knockdown with each of six shRNAs impaired proliferation and survival of four retinoblastoma cell lines (Refs. 4 and 12 and data not shown). Impaired proliferation was associated with diminished SKP2 expression and impaired S-phase entry and was partially rescued by ectopic SKP2, indicating that SKP2 is an important TR␤2 target (12).
Here, we examined the abilities of the different TR␤2 isoforms to complement endogenous TR␤2 loss. We first confirmed the prior observations in the context of a TR␤2 knockdown and complementation assay. TR␤2 knockdown and cotransduction of the BN vector caused an ϳ80% decrease in SKP2 protein but no change in SKP2 RNA (Fig. 3, A-C), confirming that TR␤2 sustains SKP2 expression at the post-transcriptional level (12). TR␤2 knockdown followed by nocodazole treatment at days 4.0 -4.5 decreased the proportion of S/G 2 /M-phase cells from 50 to 16% (Fig. 3D), confirming that TR␤2 is needed for G 1 -to-S progression. Concordantly, TR␤2 knockdown impaired Y79 cell proliferation and survival (Fig. 3E).
In cells with endogenous TR␤2 knockdown, ectopic TR␤2-WT and TR␤2-46 partially restored SKP2 levels, G 1 -S progression, S-phase entry, and proliferation, whereas TR␤2-54 failed to do so (Fig. 3, A, B, D, and E). TR␤2-WT and TR␤2-46 did not fully restore SKP2 to endogenous levels, possibly due to the inability to precisely replicate the endogenous TR␤2 levels or cell cycle-dependent expression (28). Ectopic SKP2 more fully restored SKP2 protein (Fig. 3, A and B, lane 6) but did not further restore cell cycle progression or proliferation, implying that TR␤2-46 restored sufficient SKP2 to elicit SKP2-mediated cell cycle changes. Thus, TR␤2-46 but not TR␤2-54 promoted SKP2 expression and cell cycle progression.

TR␤2-46 increases cytoplasmic SKP2 stability
Having determined that TR␤2-46 enhances expression of SKP2 protein but not SKP2 RNA, we investigated whether it does so by regulating SKP2 stability. We also examined whether ACCELERATED COMMUNICATION: TR␤2-46 promotes cell proliferation TR␤2-46 regulates SKP2 in the nucleus or in the cytoplasm as both compartments have been implicated in SKP2 function (29 -33). Through cell fractionation we found that ϳ80 -90% of SKP2 was located in the cytoplasm of vector-transduced Y79 retinoblastoma cells (Fig. 4A). After TR␤2 depletion, SKP2 declined and was seen solely in the cytoplasm. Ectopic TR␤2-46 partially restored cytoplasmic but not nuclear SKP2 (Fig. 4A) despite that, in this experiment, ectopic TR␤2-46 partially localized to the nucleus, likely due to its higher-level expression.
To assess whether TR␤2 enhanced SKP2 stability, Y79 cells were cotransduced with TR␤2 shRNAs and either the BN vec-tor or BN-TR␤2-46. On day 4, cells were treated with cycloheximide to suppress protein synthesis, and the rate of SKP2 decay was examined. In this setting, TR␤2 knockdown and SKP2 down-regulation were intentionally modest as needed to retain sufficient SKP2 to observe its half-life. As such, we observed little effect of TR␤2 knockdown on SKP2 stability, whereas ectopic TR␤2-46 stabilized and increased SKP2 expression (Fig. 4B).

Discussion
We report that retinoblastoma cells express two functionally distinct TR␤2 isoforms, here designated TR␤2-46 and TR␤2-54 according to their predicted molecular masses. Both isoforms were encoded by a THRB transcript orthologous to the well-characterized mouse THRB RefSeq isoform 2 via alternative translation initiation, with the canonical TR␤2 initiation codon used to produce TR␤2-54 and methionine 79 used to produce TR␤2-46. Methionine 79 is conserved in mice, and a TR␤2 protein of similar electrophoretic mobility appeared to be present in pituitary extracts from WT but not Tr␤2 Ϫ/Ϫ mice (25), suggesting that TR␤2-46 might be expressed in mouse pituitary. However, no faster-migrating TR␤2 was evident in mouse retinae (25,34), suggesting that species-specific mechanisms might promote TR␤2-46 expression in human cone precursors and cone precursor-derived retinoblastoma cells.
The mechanism that regulates the translation initiation of TR␤2-54 and TR␤2-46 is currently unclear. As one possibility, 5Ј-UTR sequences and upstream open reading frames (uORFs) can influence translation initiation (35)(36)(37). Retinoblastoma cell THRB cDNA had a 5Ј-UTR of at least 301 nucleotides, including multiple uORFs that are conserved in mice (Fig. 2C), suggestive of a conserved regulatory role. If the 5Ј-UTR and uORFs do indeed regulate alternative initiation then species differences in cis-acting sequences or trans-acting factors must underlie the predominant TR␤2-46 in human and TR␤2-54 in mouse retinae.

ACCELERATED COMMUNICATION: TR␤2-46 promotes cell proliferation
Translation of TR␤2-46 from methionine 79 eliminates N-terminal structures that are implicated in the enhanced transcriptional activity of TR␤2 relative to TR␤1 (16,18,38). As this enhanced transcriptional activity is implicated in regulation of the hypothalamic-pituitary-thyroid axis (39) and in long and medium wavelength cone gene expression and differentiation (6,40), TR␤2-46 seems unlikely to drive these transcriptional programs. Furthermore, the lack of TR␤2-46 in developing mouse retina (25,34) suggests that TR␤2-46 is not needed for cone development processes that are shared between mice and humans. Instead, TR␤2-46 may participate in human-specific processes such as foveagenesis or expression of a cone precursor proliferation-related program (2,3).
Our cell fractionation analyses revealed that the vast majority of TR␤2-54 and virtually all TR␤2-46 are cytoplasmically located in human retinoblastoma cell lines. This is in accord with the prior immunodetection of cytoplasmic TR␤2 in retinoblastomas and human cone precursors (2, 4, 24) and suggests that the cytoplasmic TR␤2 immunostaining was authentic. The high levels of cytoplasmic TR␤2-46 also suggest that TR␤2-46 has a cytoplasmic role. In retinoblastoma cells, TR␤2-46 but not TR␤2-54 partially sustained SKP2 expression, S-phase entry, and proliferation (Figs. 3 and 4). Although the underlying mechanism is not yet defined, it may be relevant that cytoplasmic SKP2 can be induced by AKT signaling (32,(41)(42)(43) and that cytoplasmic TR␤1 can suppress AKT signaling via inhibition of PI 3-kinase (19,20). Because TR␤2 increased SKP2 expression by antagonizing TR␤1 (12), we speculate that TR␤2-46 sustains SKP2 by opposing TR␤1-mediated inhibition of PI 3-kinase activity. The high SKP2 expression in human cone precursor cytoplasm (Fig. 4C) is consistent with the possibility that TR␤2-46 also promotes cytoplasmic SKP2 expression in the developing retina.
The TR␤2-46 -mediated up-regulation of SKP2 seems likely to contribute to retinoblastoma initiation and propagation. Indeed, SKP2 is required for the cone precursor proliferative ACCELERATED COMMUNICATION: TR␤2-46 promotes cell proliferation response to RB loss (2) as well as for retinoblastoma cell proliferation (23). Moreover, down-regulation of SKP2 in RB-depleted cells may provide an important barrier to development of RB-deficient tumors, whereas intrinsic TR␤2 expression enables SKP2 expression and RB-deficient malignancies (12). In retinoblastoma cells, TR␤2-46 increased cytoplasmic SKP2 expression along with cell proliferation. This was unexpected because SKP2 is thought to promote proliferation in part by mediating p27 degradation in the nucleus. However, SKP2 also has roles in the cytoplasm where it can ubiquitylate and activate AKT (31) and can promote degradation of the proapoptotic FOXO1/3 and the tumor-suppressive E-cadherin (32,33).
In summary, we identified a novel TR␤2-46 isoform that is highly expressed in human but not mouse cone precursors. We demonstrate that TR␤2-46 is highly expressed in cone precursorderived retinoblastoma cells and is critical to retinoblastoma cell SKP2 expression and proliferation. Thus, TR␤2-46 is a cell typespecific factor that is intrinsically expressed in the retinoblastoma cell of origin and collaborates with the cancer-initiating RB loss to enable tumorigenesis.

Cell lines and retinal tissues
Y79 and Weri-RB1 cells were from the ATCC. RB176, RB177 (4), and CHLAVC-RB43 (44) were as described. Following informed consent, fetal eyes were obtained from authorized sources with approval of the Children's Hospital Los Angeles Institutional Review Board.

Cell culture
Retinoblastoma cells were cultured in RB culture medium as described (4). Cells were synchronized at metaphase by addition of nocodazole (Sigma-Aldrich, M1404) to 100 ng/ml. Protein synthesis was blocked by addition of cycloheximide (US Biological, C8500) to 20 g/ml.

Subcellular fractionation and Western blotting
Subcellular fractionation was as described (45). For Western blotting, RB cells or minced retina was incubated with lysis buffer (10 mM Tris, pH 8.0, 140 mM NaCl, 1% Nonidet P-40, 0.1 mM EDTA with protease and phosphatase inhibitors (Roche Applied Science)) on ice for 10 min and centrifuged at 20,000 ϫ g for 10 min at 4°C, and supernatant was collected. 30 g of retinoblastoma cell protein, 10 g of retina lysate, and 5 l of molecular weight markers (Bio-Rad, 161-0317) were separated by SDS-PAGE. Antibodies to human TR␤2 amino acids 1-110 (sc-67123), SKP2 (sc-7164), GAPDH (sc-32233), and Lamin B (sc-6216) were from Santa Cruz Biotechnology, and ␣-tubulin antibody was from Sigma-Aldrich (T5168). Secondary antibodies with chemiluminescence or fluorescence signals were quantified by Imagine Studio Lite and normalized to loading controls.

TR␤2 RNA analysis
RB176 RNA was isolated, and cDNA was produced and PCRamplified as described (44)

Lentivirus production and infection
Lentiviruses were produced by transfection of 2 ϫ 10 7 293T cells similar to that described previously (4). Virus was harvested at 60 h, concentrated 50-fold by centrifugation at 25,000 rpm for 90 min, and suspended in RB medium. 500 l of concentrated virus was used to infect 5 ϫ 10 5 Y79, WERI-1, or RB177 cells in 500 l of filtered conditioned RB medium with 4 g/ml Polybrene (Sigma-Aldrich) followed by gentle pipetting 20 times. At 18 h after infection, cells were diluted in an equal volume of conditioned RB medium. Infected cells were selected starting 48 h after infection with 2 g/ml puromycin for 48 -72 h or with 200 g/ml G418 for 4 -7 days and fed every 2-3 days by replacing two-thirds of the media.

Cell cycle and cell proliferation analyses
Cells were fixed in 70% ice-cold ethanol for 1-16 h at 4°C, pelleted by centrifugation at 10,000 ϫ g for 10 s, resuspended in propidium iodide (10 g/ml in PBS and 100 g/ml RNase (Invitrogen)), incubated for 30 min at 37°C, and analyzed using a BD Canto flow cytometer with Ͼ20,000 gated events per sample. Cell cycle distributions were defined using FACSDiva version 6.1.3. Proliferation was evaluated by cell counting using a hemocytometer.

Statistical analysis
All data were from at least three independent biological repeats. One-way analysis of variance was used to determine ACCELERATED COMMUNICATION: TR␤2-46 promotes cell proliferation whether there were differences among the means of three or more groups, and then an unpaired two-tailed t test was performed to identify where the differences occurred between groups (Fig. 3, B, D, and E). A p value Ͻ0.05 was considered significant.