MYH9 binds to lncRNA gene PTCSC2 and regulates FOXE1 in the 9q22 thyroid cancer risk locus

Significance

Papillary thyroid carcinoma (PTC) is the most common endocrine cancer and displays strong heritability. So far, the most significant known predisposing variant is rs965513 in 9q22. Although a long noncoding RNA, papillary thyroid cancer susceptibility candidate 2 (PTCSC2), has been characterized in this locus, its mode of action in the carcinogenetic process is unknown. Here, we identify myosin-9 (MYH9) as a binding protein of PTCSC2 that regulates the bidirectional promoter shared by PTCSC2 and forkhead box E1 (FOXE1). PTCSC2 can rescue the promoter inhibition caused by MYH9. The p53 pathway is profoundly affected by the inhibition of FOXE1. Our study discovers fundamental roles for PTCSC2, MYH9, and FOXE1 in thyroid cancer and provides a description of the regulatory mechanism.

Abstract

A locus on chromosome 9q22 harbors a SNP (rs965513) firmly associated with risk of papillary thyroid carcinoma (PTC). The locus also comprises the forkhead box E1 (FOXE1) gene, which is implicated in thyroid development, and a long noncoding RNA (lncRNA) gene, papillary thyroid cancer susceptibility candidate 2 (PTCSC2). How these might interact is not known. Here we report that PTCSC2 binds myosin-9 (MYH9). In a bidirectional promoter shared by FOXE1 and PTCSC2, MYH9 inhibits the promoter activity in both directions. This inhibition can be reversed by PTCSC2, which acts as a suppressor. RNA knockdown of FOXE1 in primary thyroid cells profoundly interferes with the p53 pathway. We propose that the interaction between the lncRNA, its binding protein MYH9, and the coding gene FOXE1 underlies the predisposition to PTC triggered by rs965513.

Thyroid cancer is the most common endocrine malignancy. Based on the data from the National Cancer Institute for 2016, in total 64,300 new patients will be diagnosed with thyroid cancer, accounting for 3.8% of all cancer patients in the United States (https://seer.cancer.gov/statfacts/html/thyro.html). Thyroid cancer is classified into four main types: papillary, follicular, medullary, and anaplastic. Papillary thyroid carcinoma (PTC) accounts for 85–90% of all thyroid cancers (1). In efforts to explain the genomic background to PTC, genome-wide association studies (GWASs) have disclosed an SNP marker, rs965513, in 9q22.33 strongly associated with PTC in European populations [odds ratio (OR) = 1.75; P = 1.7 × 10⁻²⁷] (2, 3). The association has been observed in other populations as well (4–8). Additionally, rs965513 has been reported to be associated with levels of thyroid-related hormones, hypothyroidism, goiter, and other abnormal thyroid functions (2, 3, 9). SNP rs965513 resides in an intron of a recently detected long noncoding RNA (lncRNA), papillary thyroid cancer susceptibility candidate 2 (PTCSC2), which is located between two coding genes, DNA damage recognition and repair factor (XPA) and forkhead box E1 (FOXE1) (Fig. S1).

Fig. S1.

A diagrammatic view of the 9q22 locus. The diagram shows the lead SNP rs965513 in GWAS, the two flanking coding genes (XPA and FOXE1), and the lncRNA PTCSC2 isoforms C and D. Red arrows indicate transcriptional orientations. Blue filled boxes represent exons.

FOXE1, also known as thyroid transcription factor 2, is a single exon coding gene belonging to the forkhead/winged helix-domain protein family (10). FOXE1 knockout mice were born alive but exhibited an ectopic or completely absent thyroid gland and severe cleft palate, accompanied by up-regulated TSH and down-regulated free thyroxine hormone levels (11). FOXE1 is essential for thyroid gland development and the maintenance of thyroid differentiated status (12, 13). It also plays a role in the predisposition and development of thyroid cancer (14, 15). Moreover, somatic mutational inactivation of FOXE1 has been found in PTC, suggesting FOXE1 may contribute to tumorigenesis in a subset of thyroid cancers (16).

PTCSC2 was discovered in the region harboring rs965513 on 9q22 (15). Similar to some 25% of all known lncRNA genes (17), PTCSC2 has one unspliced isoform and several spliced isoforms, all of which display thyroid-specific expression. In PTC tumors, the risk allele [A] of rs965513 is significantly associated with low expression of both PTCSC2 and FOXE1 (15). Three enhancer elements are located in a 33-kb linkage disequilibrium (LD) block within PTCSC2. Previous studies showed that PTCSC2 is transcribed in the opposite direction of FOXE1. Exon 1 and intron 1 of PTCSC2 isoform C overlap with the FOXE1 promoter region (15). Although the promoter region shared by PTCSC2 and FOXE1 was found to be regulated via long-range looping interactions (18), the detailed mechanisms and their bearing on thyroid cancer need to be explored.

In the present study, we identified myosin-9 (MYH9) as an lncRNA binding protein that targets the FOXE1 promoter region through interactions with PTCSC2 and performs its regulatory function in thyroid cancer via downstream pathways. These findings provide a better defined description of the complex mechanisms involved in the 9q22 thyroid cancer locus.

Results

Identification of MYH9 as an LncRNA Binding Protein.

To begin to unravel the underlying mechanisms by identifying proteins potentially interacting with PTCSC2, we performed biotin-labeled RNA pull-down assays with protein lysate from normal thyroid tissue. Two PTCSC2 isoforms—isoform C (1,947 nt) and isoform D (1,804 nt)—were used as baits by generating both sense and antisense (negative control) RNA probes for each of them (Fig. S1). As a result, a strand-specific binding protein was discovered between 150 KD and 250 KD in size (Fig. 1A and Fig. S2). This protein was identified as MYH9, being 226 KD in size by mass spectrometry (MS) (Fig. 1B). Another protein band that was noticed between 37 KD and 50 KD in size was identified as beta-actin (ACTB) (Fig. 1A and Figs. S2 and S3). As MYH9 is an ACTB binding partner participating in important cellular processes (19), ACTB was not studied further here.

Fig. 1.

Identification of MYH9 as a PTCSC2 binding protein. (A) RNA pull-down experiment with nontumorous thyroid tissue extract. Biotin pull-down assays followed by SDS/PAGE separation were used to isolate the binding protein of PTCSC2 isoform C. Antisense RNA of PTCSC2 isoform C was used as the negative control. The arrows indicate the binding protein bands ∼226 KD and 42 KD in size, respectively. (B) Information for the identification of MYH9 by MS. The full length of MYH9 protein is 1,960 amino acids, as shown in the lower diagram. Blue boxes indicate the peptides identified by MS analysis. (C) qRT-PCR detection of the indicated RNAs retrieved by MYH9 antibody (RIP assay) in the BCPAP cell line with stable PTCSC2 isoform C overexpression. IgG was used as a negative control. The relative fold change was normalized to the IgG control. **P < 0.01. Student’s t test.

Fig. S2.

Identification of MYH9 as a PTCSC2 isoform D binding protein. (A) RNA pull-down experiment with nontumorous thyroid tissue extract. Biotin pull-down assays followed by SDS/PAGE separation were used to isolate the protein binding PTCSC2 isoform D. Antisense RNA of PTCSC2 isoform D was used as the negative control. The arrows indicate the binding protein bands ∼226 KD and 42 KD in size, respectively. (B) Information for the identification of MYH9 by MS.

Fig. S3.

The detailed information of MS analysis for ACTB protein identified in the RNA pull-down assay. Information for the identification of beta-actin by MS.

The interaction between PTCSC2-isoform C and MYH9 was confirmed by RNA immunoprecipitation (RIP) assay in BCPAP cells with PTCSC2-isoform C stable overexpression (Fig. 1C). PTCSC2 was readily enriched by MYH9 antibody compared with IgG-negative controls (>sixfold), and this enrichment was significantly higher than in the 18S RNA controls (P < 0.01), demonstrating the specificity of the assay.

MYH9 Binds to the FOXE1 Promoter Region.

To determine whether the PTCSC2 binding protein MYH9 is involved in the transcriptional expression of the FOXE1 gene, four overlapping regions (R1, R2, R3, and R4) that cover the transcription factor enriched region (from ENCODE ChIP sequencing data) in the 5′UTR of FOXE1 were chosen for ChIP analysis using MYH9 antibody (Fig. 2A). Of note, the previously reported thyroid cancer risk-associated SNP rs1867277 (14) is located in the R3 genomic region. In the KTC1 cell line, significant enrichment was found in the R1 and R3 regions (Fig. 2B). The enrichment was confirmed in nontumorous thyroid tissue (Fig. 2C), suggesting the DNA binding activity of MYH9 is in the FOXE1 promoter, perhaps more specifically in regions R1 and R3.

Fig. 2.

Validation of the MYH9 binding site in the FOXE1 promoter region by ChIP assay. (A) Schematic view of a 900-bp transcription factor enrichment region in the FOXE1 promoter region. Genomic coordinate, chr9:100,615,431–100,616,330 (GRC37/hg19). DNase I hypersensitive sites, CpG islands, and ENCODE data were obtained from the UCSC Genome Browser. Red bars indicate the tested regions in ChIP assays. (B) ChIP assay using KTC1 cell line. The results represent four independent experiments, each in duplicates. Results are shown as means ± SD. *P < 0.05. Student’s t test. (C) ChIP assay using nontumorous thyroid tissue sample. The relative abundance was normalized to the input control.

MYH9 Functions as a Suppressor in the Promoter Region of PTCSC2 and FOXE1 with Bidirectional Activity.

To investigate the role of MYH9 binding to the FOXE1 promoter region, luciferase assays were performed in cells transiently transfected with FOXE1 promoter or empty constructs (without promoter), together with MYH9 expression or empty expression vectors.

As PTCSC2 and FOXE1 are transcribed in opposite directions (15) and there is only 174 bp between their transcription start sites (TSSs), we postulated that they are transcribed by the same promoter with bidirectional activity. To test this hypothesis, a shorter (1,607 bp) and a longer (2,520 bp) fragment cloned from the FOXE1 promoter region were used for luciferase reporter assay in the BCPAP cell line with both forward (Promoter-S and Promoter-L) and inverted (Promoter-Inv-S and Promoter-Inv-L) orientation, as indicated (Fig. 3A). Although the inverted promoters showed lower activity compared with their corresponding forward ones, the fragment with both orientations exhibited a significantly higher level of promoter activity in comparison with the no promoter control. Moreover, cotransfection with the MYH9 expression vector showed that MYH9 can significantly inhibit the luciferase activity in both the forward and inverted promoter fragments. That MYH9 functions as a suppressor was disclosed in both the shorter and longer promoter fragments (Fig. 3 B and C).

Fig. 3.

Effect of MYH9 on FOXE1 promoter with bidirectional activity. (A) Schematic view of the vectors used in luciferase assays. Red boxes indicate the UTR regions of FOXE1; the yellow box indicates the FOXE1 coding sequences; blue boxes indicate the exons of PTCSC2; the thick gray line indicates the intron of PTCSC2; the thin black line indicates the intergenic genomic region. All of the position numbers are labeled relative to the FOXE1 TSS as +1. Arrows represent the fragment direction in the constructs according to the position in the genome. Dual reporter luciferase assays in BCPAP cells cotransfected with constructs containing a short promoter fragment (B) or long promoter fragment (C) having either forward or inverted orientation, with or without MYH9 overexpression vector. Values are presented as fold of the no promoter control treated with empty expression vector. Results are shown as means ± SD of four independent experiments, each in four replicates. *P < 0.05; **P < 0.01; ***P < 0.001. Student’s t test.

Thus, MYH9 had an impact on the FOXE1 gene, but its role in the presence of lncRNA PTCSC2 was not known. We transiently cotransfected the PTCSC2 expression vector in the BCPAP cell line in the presence or absence of the MYH9 expression construct. Because there was no significant difference in promoter activity between the two fragments of different sizes, only the longer promoter sets were used. Interestingly, an increase in promoter luciferase activity was observed with cotransfection of the PTCSC2 expression vector, irrespective of MYH9 overexpression. The effect was significant on the forward Promoter-L, whereas it was not significant on the inverted Promoter-Inv-L, suggesting a rescuing effect on the inhibition caused by MYH9 (Fig. 4 A and B).

Fig. 4.

Effect of PTCSC2 on FOXE1 bidirectional promoter. Dual reporter luciferase assay using the long promoter constructs with forward (A) or inverted (B) orientations cotransfected with PTCSC2, MYH9, or empty expression vectors. The values at the bottom of each column represent the portion of the total expression vectors. Results are shown as means ± SD of four independent experiments, each in four replicates. *P < 0.05. Student’s t test. (C) qPCR detection of endogenous FOXE1 in the BCPAP cell line and (D) the TPC1 cell line transiently transfected with PTCSC2, MYH9, or empty expression vectors as indicated (n = 3). All values were normalized with the values of the corresponding empty vector-transfected groups. Results are shown as means ± SD. *P < 0.05; NS, not significant. Student’s t test.

To further validate the roles of MYH9 and PTCSC2 in regulating the endogenous FOXE1 promoter, a transient transfection assay with MYH9 and PTCSC2 expression vectors was performed in BCPAP and TPC1 cells. The expression of the endogenous FOXE1 gene showed a significant decrease in the presence of MYH9 overexpression, which is consistent with the role of MYH9 as an inhibitor in our luciferase assay. This inhibitory effect becomes nonsignificant in the presence of PTCSC2, indicating that FOXE1 is regulated by PTCSC2 and MYH9 differentially (Fig. 4 C and D). When we examined the PTCSC2 transcript levels and protein levels of MYH9 and FOXE1 in 6 thyroid cancer cell lines and in nontumorous thyroid tissue, we did not notice a correlation among the expression levels of these three genes (Fig. S4).

Fig. S4.

Expression of MYH9 and FOXE1 in thyroid cancer cell lines and nontumorous thyroid tissue. The protein levels of MYH9 and FOXE1 were detected by Western blotting using ACTB as the loading control. Expression of PTCSC2 spliced isoforms and GAPDH as detected by RT-PCR was described previously in ref. 15.

SNP rs1867277 is in partial LD with rs965513 and is reported as a functional variant in the regulation of FOXE1 expression and confers thyroid cancer susceptibility (14). As rs1867277 is located in the MYH9 interacting region R3 (Fig. 2A), we tested the regulatory effects of MYH9 and PTCSC2 on the FOXE1 promoter with either the wild-type allele (G) or the thyroid cancer risk allele (A). BCPAP cells were transiently transfected with either MYH9 or both MYH9 and PTCSC2 constructs using promoters with opposite orientations. The promoter containing the A allele exhibited significantly lower activity than the promoter with the G allele, and a similar difference was found in both forward and inverted promoters (Fig. S5).

Fig. S5.

Transcriptional activity of FOXE1 promoter with different alleles of rs1867277. Dual reporter luciferase assay using long promoter constructs with either G or A allele of rs1867277 and forward (Left) or inverted (Right) orientations cotransfected with MYH9, PTCSC2, or empty expression vectors. All values were normalized with the values of the corresponding groups using promoter plasmids containing the G allele. Results were shown as means ± SD of four independent experiments, each in four replicates. *P < 0.05; ***P < 0.001. Student’s t test.

FOXE1 Is Involved in the P53 Pathway in Human Primary Thyroid Cells.

FOXE1 is an important transcriptional regulator in the physiological functions of the thyroid as well as in thyroid cancer. Lower expression level of FOXE1 is associated with PTC risk in the presence of the risk allele of SNP rs965513 (15). To identify genes involved in FOXE1-mediated signaling pathways and transcriptional regulation in thyroid, knockdown of FOXE1 was performed with siRNAs in nontumorous human primary thyroid cells (20). Nontumorous human primary thyroid cells from the same patient sample transfected with scrambled siRNAs were used as the control. Knockdown of FOXE1 resulted in numerous changes in gene expression determined by RNA deep sequencing (fold change > 1.5, P < 0.001). Of the dysregulated genes (total N = 107; up-regulated n = 80), 89.7% were coding genes (Table S1). Biological functional analysis using Ingenuity Pathway Analysis (IPA) software showed that cancer was in the top category of “Diseases and Disorders” with the second lowest P value (Fig. S6). The top five categories of “Molecular and Cellular Functions”—namely, cellular movement, cell death and survival, cellular development, cellular growth and proliferation, and cell signaling—suggested its possible relevance to pathways related to cell apoptosis, migration, or invasion (Table 1). These effects have been widely reported as the major effects of the well-known p53 signaling pathway (21, 22). Of the top 25 dysregulated genes, the thrombospondin 1 (THBS1, also known as TSP-1) and insulin-like growth factor binding protein 3 (IGFBP3) genes were found to be up-regulated in FOXE1 knockdown cells (Fig. 5A and Table S1). This inverse relationship was subsequently confirmed by quantitative RT-PCR (qRT-PCR) in primary thyroid cells (Fig. 5B) and also in the BCPAP and TPC1 cell lines (Fig. S7). We also analyzed gene expression data from 59 pairs of tumor/nontumor PTC samples available in the TCGA data portal (https://tcga-data.nci.nih.gov/docs/publications/tcga). In this analysis, the expression of FOXE1 was significantly (P = 2.02 × 10⁻⁰⁵) higher in nontumorous tissue, whereas the expression of both THBS1 and IGFBP3 are significantly higher in tumorous tissue (P = 1.73 × 10⁻⁰² and P = 4.19 × 10⁻⁰⁵, respectively) (Table S2). This is consistent with our data and further supports the putative role of FOXE1 as a suppressor of THBS1 and IGFBP3 in thyroid. Both THBS1 and IGFBP3 are important members of the p53 pathway involved in apoptosis, inhibition of the IGF1/mTOR pathway, and inhibition of angiogenesis and metastasis, suggesting interactions between FOXE1 and the p53 pathway in thyroid cells (Fig. 5C).

Table S1.

Top 107 dysregulated genes by FOXE1 knockdown in thyroid primary cell cultures

Gene name	P value	Fold change
GDF6	1.43E–20	2.435
OASL	4.58E–15	2.101
ACTC1	4.26E–13	2.099
AQP3	2.78E–09	1.925
FOXE1	1.14E–14	−1.908
IGFBP3	1.65E–12	1.894
SYTL2	1.08E–15	1.893
STC1	3.31E–17	1.883
PLAU	1.93E–14	1.878
IVL	1.23E–08	1.843
SALRNA2	6.07E–08	−1.824
RAET1L	7.85E−10	1.807
SLC19A2	3.20E–14	1.798
ELF3	7.60E−14	1.793
CORO6	4.56E–13	−1.782
ADM	1.00E−10	1.778
KRT17	1.88E–10	1.774
IL11	7.06E−09	1.769
TGM2	2.26E–15	1.754
NGFR	2.24E–07	1.751
ALOX15B	1.23E–11	−1.745
TREM1	1.05E–09	1.744
CNN1	6.15E–08	1.731
THBS1	1.74E–13	1.717
TACSTD2	3.59E–10	1.707
RTN4RL2	3.11E–07	−1.706
AJAP1	5.84E–13	1.703
P4HA3	6.19E–09	1.698
BRINP1	4.66E–07	1.692
SH3TC2	4.16E–09	1.689
SYBU	4.42E–11	1.688
PORCN	2.40E–10	1.683
AC096669.3	1.21E–06	−1.681
RGS4	1.07E–06	1.681
RP11-349F21.5	3.09E–06	−1.677
PLEKHD1	5.07E–10	−1.662
SLC5A5	1.64E–06	−1.653
PFKFB4	9.49E–10	1.639
TUBA4A	5.77E–10	1.631
MIR7848	1.11E–06	−1.630
CALCB	6.86E–08	1.625
ZNF114	1.38E–08	1.611
RASSF4	4.63E–10	1.610
RP11-362F19.1	7.87E–08	1.603
KHDRBS2	1.56E–06	−1.603
DIRAS3	5.83E–06	1.598
RP11-349F21.2	1.69E–05	−1.598
EDNRA	4.52E–08	1.591
LMCD1	1.08E–09	1.587
PPP1R1A	7.95E–07	1.585
KRT19	2.44E–10	1.580
SLC38A5	5.24E–09	1.570
SFN	6.22E–08	1.569
CXCL8	2.31E–06	1.569
HTRA3	1.74E–07	−1.569
RRAD	3.77E–07	1.569
RP11-626E13.1	3.73E–06	−1.569
SSX2	5.18E–05	1.568
RP11-486B10.4	2.27E–05	−1.567
SLA	8.89E–07	−1.566
CCL2	4.24E–05	1.564
C3orf36	1.89E–06	1.560
FAM110B	3.70E–09	1.559
NPPC	6.27E–05	1.559
PPP1R3C	1.09E–05	1.558
PDE4A	1.43E–05	−1.558
CNN3	6.67E–08	1.557
ADAMTS9	7.14E–08	1.557
PLCXD3	6.62E–05	1.557
SGK1	2.51E–08	1.555
IER3	2.63E–07	1.554
SHC4	3.95E–05	1.554
IPCEF1	8.07E–06	−1.551
SGCD	1.53E–08	−1.547
SLC43A3	2.23E–09	−1.547
TNC	5.65E–10	1.547
GBP1	1.56E–06	1.546
EDN1	4.25E–05	1.544
MAMDC2	1.07E–06	1.541
C1orf116	6.43E–09	1.540
KCNN4	2.98E–05	1.540
RAB36	4.48E–07	−1.540
ISM1	1.07E–06	−1.537
TNFRSF19	3.30E–07	1.536
PRKAR2B	6.80E–06	1.530
TGFA	4.70E–08	1.529
RSAD2	5.86E–06	1.527
SSX2B	1.48E–04	1.524
KRT83	3.56E–06	1.523
PHLDB2	7.77E–07	1.521
PTCHD1	3.76E–05	−1.519
HGF	1.70E–04	−1.518
NPBWR1	1.84E–04	1.515
TNFRSF11B	6.83E–06	1.515
LMLN	9.49E–06	−1.515
MDFI	8.24E–06	1.515
RP11-506K6.4	1.95E–04	−1.512
DCBLD2	1.03E–07	1.512
TIMP4	5.52E–05	1.510
EHF	1.95E–04	1.506
Y_RNA	2.32E–04	−1.505
WWC1	8.29E–10	1.505
CRABP2	2.52E–05	1.505
DPYSL4	3.94E–05	1.502
NTNG1	1.09E–04	1.502
TNNC1	1.08E–04	1.502
LINC00511	1.60E–04	1.501

Fig. S6.

The top 10 key biological functional groups predicted by dysregulated genes in FOXE1 knockdown thyroid primary cells by IPA analysis.

Table 1.

Molecular and cellular functions of the dysregulated genes in FOXE1 knockdown cells by IPA analysis

Function	P value	Molecules
Cellular movement	1.30 × 10−03–7.65 × 10−10	42
Cell death and survival	1.27 × 10−03–8.74 × 10−10	46
Cellular development	1.22 × 10−03–1.20 × 10−08	53
Cellular growth and proliferation	1.13 × 10−03–1.20 × 10−08	47
Cell signaling	9.58 × 10−04–6.75 × 10−08	19

Fig. 5.

Gene expression profile of FOXE1 knockdown in thyroid primary cells indicating its involvement in the p53 pathway. (A) Gene expression differences of the top 25 genes caused by FOXE1 knockdown in primary thyroid cells on day 5 of culture. The expression is plotted with heat-map color scale using relative expression fold change (FOXE1 knockdown culture vs. control culture) (fold change > 1.5, P < 0.001). Exp. 1, Exp. 2, and Exp. 3 represent three independent primary culture experiments derived from different patient samples. (B) Confirmation by qRT-PCR of two up-regulated genes, IGFBP3 (Left) and THBS1 (Right), in FOXE1 knockdown primary cells. si-Control represents the primary cell sample treated with scrambled siRNA control. Results are shown as means ± SD of three independent experiments, each in three replicates. All values were normalized with the values of the corresponding negative control siRNA-treated groups. *P < 0.05; ***P < 0.001. Student’s t test. (C) Part of the p53 signaling pathway including typical gene members and their functions from Kyoto Encyclopedia of Genes and Genomes (KEGG). THBS1 is labeled in a purple ellipse, and IGFBP3 is labeled in two red ellipses. (D) Cell viability changes of BCPAP cell line at three time points (24, 48, and 72 h) with FOXE1 knockdown. For each time point, experiments were performed in four replicates of three biological replicates. All values were normalized with the values of the corresponding negative control siRNA-treated groups. (E) Example plots of cell apoptosis assays in BCPAP cell line treated with FOXE1 siRNA or negative control siRNA. Cells stained with only annexin V were evaluated as being in early apoptosis stage (lower right quadrant); cells stained with both annexin V and PI (propidium iodide) were evaluated as being in late apoptosis stage (upper right quadrant).

Fig. S7.

qRT-PCR of FOXE1, THBS1, and IGFBP3 in FOXE1 knockdown cell lines. Expression levels of BCPAP cell line (Left) or TPC1 cell line (Right) were detected after treatment with FOXE1 siRNA or negative control siRNA for 24 h. Results are shown as means ± SD of three independent experiments, each in three replicates. All values were normalized with the values of the corresponding negative control siRNA-treated groups. *P < 0.05; **P < 0.01; ***P < 0.001. Student’s t test.

Table S2.

Differential gene expression of 59 tumorous (T)–nontumorous (N) PTC tissue pairs from TCGA database for FOXE1, IGFBP3, and THBS1

Gene name	P value*	Fold change†
FOXE1	2.02E–05	−1.504
IGFBP3	4.19E−05	1.668
THBS1	1.73E–02	1.606

^*P values were obtained by paired t test.

^†The values represent the median fold changes of tumorous versus nontumorous tissue pairs for each gene.

To further assess the function of FOXE1 in thyroid cells, we performed cell viability and apoptosis assays in FOXE1 knockdown BCPAP cells. The down-regulation of FOXE1 led to reduced cell viability in CellTiter-Glo assay (Fig. 5D). Consistently, apoptosis assay with flow cytometric analysis indicated that increases of apoptotic cells at both early and late apoptosis stages were found in the FOXE1 knockdown cells (Fig. 5E).

Discussion

LncRNAs are important regulators of tissue physiology and disease processes including cancer. The regulatory effectiveness of lncRNAs is dependent on their expression. Many lncRNAs show tissue-specific expression patterns. Several thyroid tissue- and cancer-associated lncRNAs were recently discovered (23). Only a few lncRNAs related to thyroid and thyroid cancer have been well characterized, including PTCSC2, PTCSC3, and NAMA (15, 24, 25), but even in these cases, the relevant molecular mechanisms are not known. It is of special interest that these lncRNAs are involved in the genetic predisposition to thyroid cancer (germ-line involvement). Therefore, their impact on nontumorous thyroid tissue needs to be explored, which we here have endeavored to accomplish. We identified MYH9 as a noncoding RNA binding partner, which had not been reported before.

MYH9 protein is a member of the nonmuscle myosin II (NMII) group, which belongs to the myosin II subfamily (26). As a subunit of myosin II heavy chains, MYH9 takes part in the generation of cell polarity, cell migration, cell–cell adhesion processes, and maintaining cytoskeleton structure by binding to actin filaments (19). Recently, additional functions of MYH9 have been reported. For example, MYH9 affects the expression of PAX5 by interacting with Thy28 (27), and it can activate AKT through RAC1 and PAK1 (28), suggesting a role in gene regulation. In 2014, a role for MYH9 as a tumor suppressor gene was discovered in squamous cell carcinomas (SCCs). This report implicated MYH9 in tumor development by regulating posttranscriptional p53 stabilization, but the underlying mechanism is still unknown (29). Our results are consistent with this finding and provide further evidence for a regulatory model by which a specific noncoding RNA, PTCSC2, and MYH9 work together by regulating FOXE1 promoter activity. Knockdown of FOXE1 in primary human thyroid cells revealed that FOXE1 can regulate events in the p53 pathways.

As a common occurrence, bidirectional transcription of two protein coding genes has been discovered and studied during the past decades (30, 31). More recently, many noncoding RNAs that are transcribed in the proximity of coding genes and driven by the same bidirectional promoter have been reported (32, 33). Some noncoding transcripts exert repression of their nearby coding genes by competing for the same polymerases and accessory factors or by transcriptional interference or histone modification (34–36). A number of highly expressed antisense transcripts derived from bidirectional TSSs are able to enhance the expression of the corresponding protein coding gene in a tissue-specific manner (37). The PTCSC2–FOXE1 pair in our study further supports this regulatory model, which can be directed by the same inhibitor. Considering the complexity of the 9q22 region including nearby CpG islands (38), it is reasonable to assume that other transcription factors and some epigenetic modifications might be involved in the regulation of this bidirectional promoter, in addition to MYH9 and PTCSC2. It will be of interest to learn whether other regulatory factors have any impact on MYH9, or vice versa.

Cell line models were used in most previous reports on FOXE1 in the thyroid (39, 40). Because several thyroid markers including lncRNAs lose their expression in most cell lines (41), we consider thyroid primary cell culture to be a superior model for pathway and network analysis. In this study, we use human thyroid primary cell cultures that were recently established (20).Two important members of the p53 pathway, THBS1 and IGFBP3, have been reported to be involved in various cancer types. The regulation of tumor growth and metastasis by THBS1 has been widely described (42). THBS1 is regulated by BRAF^V600E in thyroid cancer cells. The silencing of THBS1 results in decreased cell proliferation, adhesion, migration, and invasion (43). Moreover, THBS1 silencing can also cause changes in the levels of integrin receptors and anaplastic thyroid cancer cell morphology (44). THBS1 promotes human follicular thyroid carcinoma cell invasion mainly through up-regulation of urokinase plasminogen activator (PLAU) (45). This is consistent with our RNA-seq results, as PLAU is also one of the top 25 dysregulated genes in FOXE1 knockdown primary thyroid cells (Fig. 5A and Table S1). IGFBP3 inhibits tumor growth by promoting apoptosis and inhibiting cell proliferation in some cancer types, but in other circumstances, it increases cell survival and stimulates proliferation (46). Our data further emphasize the roles played by these two genes (THBS1 and IGFBP3) in thyroid cancer.

In conclusion, our study addresses the mechanisms and provides a biological characterization of the widely reported GWAS locus 9q22 in thyroid cancer. We propose MYH9 as a binding protein of thyroid-specific noncoding RNA that may underlie the predisposition to thyroid cancer.

Materials and Methods

The study was approved by the Institutional Review Board at The Ohio State University (OSU), and all subjects gave written informed consent before participation.

Cell Lines and Thyroid Tissue Samples.

The human thyroid carcinoma cell lines used in this study were incubated in antibiotic-free RPMI medium 1640 (KTC1, BCPAP, C643, SW1736) or DMEM (TPC1, FTC133) supplemented with 10% (vol/vol) FBS (Gibco) at 37 °C in humidified air with 5% (vol/vol) CO₂.Thyroid nontumorous samples were snap-frozen in liquid nitrogen and kept at –80 °C after being obtained from patients with PTC during surgery. All cases were histologically diagnosed as PTC; clinical information on these samples will be made available on request.

In Vitro RNA Probe Synthesis and RNA Pull-Down.

In vitro RNA synthesis procedures using MEGAscript T7 Transcription Kit for target probe and MEGAscript SP6 Transcription Kit for antisense control probe (Life Technologies) were performed according to the manufacturers’ instructions. The whole-protein lysate from the human nontumorous thyroid sample was extracted using T-PER Tissue Protein Extraction Reagent (Thermo Fisher Scientific). Protein concentration was determined by using the Bio-Rad Protein Assay Dye Reagent Concentrate (Bio-Rad) kit. RNA pull-down assay was performed using Pierce Magnetic RNA-Protein Pull-Down Kit (Thermo Fisher Scientific) according to the standard instructions. A detailed description can be found in SI Materials and Methods.

RIP Assay.

In total, 2 × 10⁶ BCPAP cells with PTCSC2 isoform C overexpression were used for RIP lysis preparation using Harsh Lysis Buffer of Imprint RIP Kit (Sigma-Aldrich) following the instructions. Briefly, 5 µg antibody of rabbit anti-MYH9 (Santa Cruz, sc-98978) or rabbit IgG (Sigma, I5006-1MG) was used to incubate with the cell lysate in 1 mL RIP wash buffer with Protease Inhibitor Mixture (Roche) and also RNase Inhibitor overnight at 4 °C after ligation to Maganetic Beads. The RIP products were harvested by washing with 1 mL washing buffer up to five times. Finally, the washed product resuspended in 200 µL buffer was subjected to the purification step using TRIzol reagent (Life Technologies). The quantification of target RNA in the final RIP products was determined by real-time qRT-PCR using the Taqman method.

ChIP Assay.

A detailed description of the chromatin immunoprecipitation (ChIP) assays to determine the degree of MYH9 enrichment on the KTC1 cell line and frozen thyroid tissue samples can be found in SI Materials and Methods. The primer sequences are provided in Table S3.

Table S3.

PCR primer sequences

Gene or region name	Forward	Reverse
Primers for qRT-PCR using SYBR Green kit
THBS1	CCTGTGATGATGACGATGA	CTGATCTGGGTTGTGGTTGTA
IGFBP-3	CCATGACTGAGGAAAGGAGCTC	TGCAGCAGGGCAGAGTCTC
Primers for q-PCR in ChIP assay
R1	GCCCAGCGCCAGTACTAACT	CTGTGGTGCCCGCTAGTTTA
R2	CTAAACTAGCGGGCACCACA	CGTGACCGGGACTGGACT
R3	CTTCAGCCGGAGACCAGAGT	ACAGAGGCTCGGGAGTGAC
R4	TCGGCTAGCGGGTCACTC	GAGAGCTCAGGGGATCGTC

Transfection and Dual Luciferase Reporter Assay.

For the luciferase reporter assay, BCPAP cells were transiently transfected with reporter plasmid using Lipofectamine 3000 Reagent (Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, cells were seeded in a 24-well plate at 0.7 × 10⁵ cells per well and cultured overnight. The following day, 250 ng luciferase reporter plasmids and 250 ng effector plasmids were cotransfected along with 1.25 ng Renilla plasmid pRL-TK (Promega) as a control for each well. Cells were lysed with 100 µL passive lysis buffer (Promega) for luciferase activity analysis using the Dual-Luciferase Reporter Assay System (Promega) 24 h after transfection. A 20-μL aliquot of cell lysate was then assayed for luciferase activity using GloMax 96 Microplate Luminometer (Promega).

Cell Viability Assay and Cell Apoptosis Assay.

Cell viability in the BCPAP cell line transfected with FOXE1 siRNA or negative control siRNA at different time points (24, 48, and 72 h) was analyzed by using the CellTiter-Glo assay (Promega). The numbers of viable cells were represented by the luminescent signals, which were obtained by using GloMax 96 Microplate Luminometer (Promega). Experiments were performed according to the manufacturer’s instructions. Apoptotic cell death in the BCPAP cell line transfected with FOXE1 siRNA or negative control siRNA after 72 h was measured using the FITC Annexin V Apoptosis Detection Kit (BD Biosciences) by flow cytometry according to the manufacturer’s instructions. Three biological replicates were performed.

SI Materials and Methods

In Vitro RNA Probe Synthesis and RNA Pull-Down.

In vitro RNA synthesis procedures using MEGAscript T7 Transcription Kit for target probe and MEGAscript SP6 Transcription Kit for antisense control probe (Life Technologies) were performed according to the manufacturers’ instructions. Template DNAs were prepared using T7/SP6 primer and specific primers by PCR of cDNA clones containing PTCSC2 isoform C or isoform D transcripts, respectively. This was followed by purification with QIAquick Gel Extraction Kit (QIAGEN) after the agarose gel band had been dissolved in RNase free water. Sequences were verified by Sanger sequencing. Synthesized RNAs were analyzed qualitatively and quantitatively by electrophoresis and Nanodrop 2000 and then stored at –80 °C.

Primers for Target RNA probe are as follows: forward (T7), TAATACGACTCACTATAGGGACAACGTCA-GGGCGGGGAGGGCGC; reverse, ATGAATTAAGAGTCCTTTATTAGC. Primers for antisense control RNA probe are as follows: forward (Sp6), ATTTAGGTGACACTATAGAAATGAATTAA-GAGTCCTTTATTAGC; reverse, ACAACGTCAGGGCGGGGAGGGCGC.

The whole protein lysate from human nontumorous thyroid sample was extracted using T-PER Tissue Protein Extraction Reagent (Thermo Fisher Scientific). Protein concentration was determined by using Bio-Rad Protein Assay Dye Reagent Concentrate (Bio-Rad) kit. RNA pull-down assay was performed using Pierce Magnetic RNA-Protein Pull-Down Kit (Thermo Fisher Scientific) according to the standard instructions. Briefly, the target RNA and antisense control RNA were labeled with Biotin at the 3′ end and purified using Pierce RNA 3′ End Desthiobiotinylation Kit (Thermo Fisher Scientific). Labeled RNA probe (50 pmol) was used for the binding to Streptavidin Magnetic Beads after incubation in 1× RNA Capture Buffer for 30 min at room temperature. In total, 200 µg protein was used for the subsequent protein binding step in protein–RNA binding buffer for 150 min at 4 °C with agitation. The final RNA–magnetic beads–protein complex was washed three times with wash buffer, whereafter 12 µL elution buffer was added to retrieve the pull-down protein products. The retrieved proteins were resolved in gradient gel electrophoresis followed by MS identification.

ChIP Assay.

ChIP assays to determine the degree of MYH9 enrichment were performed using the Magna ChIP A/G Chromatin Immunoprecipitation Kit (17-10085, EMD Millipore) on KTC1 cells according to the manufacturer’s instructions. Briefly, chromatin was cross-linked with 1% formaldehyde for 10 min at room temperature. After sonication, chromatin was immunoprecipitated with rabbit anti-MYH9 antibody (sc-98978X, Santa Cruz Biotechnology) or IgG (sc-2027X, Santa Cruz Biotechnology) at 4 °C overnight. The protein/DNA complexes were eluted from the magnetic beads after standard washing steps. The cross-links were reversed by incubating at 62 °C for 2 h and 95 °C for 10 min. Final DNA products were purified by using QIAquick PCR Purification Kit (QIAGEN). Then, qPCR assays were performed by using the purified DNA as template with primers covering the transcription factor-enriched region of the FOXE1 promoter (Fig. 2A).

ChIP assays on frozen thyroid tissue samples were performed as follows: Briefly, ∼50 mg of frozen tissue was minced into small pieces (∼2 mm) using sterile blades. Protein/DNA cross-linking was performed by incubating with formaldehyde at 1% concentration for 10 min at room temperature. After washing twice with PBS, fixed tissues were homogenized. Cell pellets were resuspended after centrifugation. Then, sonication and the following steps were carried out as previously described (18).

Luciferase Plasmid Constructs.

The promoter region of FOXE1 was PCR-amplified from human BAC clone RP11-746L3 and cloned into the KpnI and SacI sites of the PGL4.10 vector (Promega). Two sets of plasmids were constructed with a long fragment (–2,062 to +1 of the FOXE1 upstream regulatory region from TSS and +1 to +458 of 5′UTR) and a short fragment from (–1,149 to +1 of the FOXE1 upstream regulatory region from TSS and +1 to +458 of 5′UTR). Another two sets of plasmids with inverted promoter regions were made from their corresponding forward plasmids by end blunting of the insertion and religation into the vector. Orientation-specific clones were screened by PCR, and all of the constructs were validated by Sanger sequencing. The expression vectors were pCMV-mCherry-MHC-IIA (Addgene plasmid 35687, a gift from Venkaiah Betapudi, Case Western Reserve University, Cleveland) (47), pcDNA3-PTCSC2 (15), and pcDNA3 (Thermo Fisher Scientific), which was used as the empty vector control.

Quantitative Real-Time PCR Assay.

Real-time qRT-PCR assay was performed in three biological replicates on ABI Prism 7900 HT Sequence Detection System (Applied Biosystems) according to the manufacturer’s protocol. The Taqman assays were carried out using Taqman probe sets for PTCSC2 Isoform C (15), FOXE1 (Life Technologies, Hs00916085_sl), and 18S (Life Technologies, 4333760T) with TaqMan Fast Universal PCR Master Mix (Thermo Fisher Scientific). THBS1, IGFBP-3, and all of the primer sets used in ChIP assay (Table S3) were detected by Fast SYBR Green Master Mix kit (Thermo Fisher Scientific).

Human Primary Thyroid Cell Culture and Nucleofection.

Human primary thyroid cells were cultured in customized 6H medium as previously described (20). Briefly, fresh nontumorous thyroid tissue samples (0.3∼1.5 g) were obtained from patients with PTC by surgery. The tissue was immediately dissected into fragments as small as possible using a sterile razor blade in a cell culture hood. After one wash in Hanks’ Balanced Salt Solution (Life Technologies), the tissue fragments were transferred to 0.25% trypsin solution for an overnight digestion. On the second day, the fragments were digested with 1% trypsin (Life Technologies) and 0.35% collagenase 4 (Worthington Biochemical) solution for 90 min at 37 °C. The digested material was filtered through nylon mesh (100 µm, FALCON). After centrifugation at 1,000 × g for 5 min, the supernatant was discarded and 1 mL red blood cell lysing buffer (Sigma) was added for 2 min to eliminate the blood cells. The cells were washed twice with Hanks’ solution and centrifuged at 1,000 × g for 5 min. Finally, the cells were counted using a TC20 Automated Cell Counter (Bio-Rad) and seeded to a density of 10⁵∼10⁶ cells per well on six-well plates.

Human FOXE1 siRNA (ON-TARGET plus SMART pool) and negative control siRNA (ON-TARGET plus control pool) were purchased from Dharmacon. Human thyroid primary cells were cultured in 6H medium for 5 d until they had reached 80∼90% confluency. Then, primary cells were electroporated by NucleofectorII device (Amaxa) with 75 pmol siRNA in 100 μL of Basic Nucleofector Medium for Primary Mammalian Epithelial Cells (Lonza, VPI-1005) for each well of the six-well plates. Cells were then resuspended in 6H medium, incubated for 24 h, and used for further analysis.

RNA-Seq Sample Preparation and Detection.

The total RNA samples for RNA-seq were extracted by TRIzol reagent (Invitrogen) and then treated by DNase-I (Ambion) to eliminate DNA contamination. RNA concentration was determined by using Qubit 2.0 Fluorometer (Agilent Technologies) with an RNA HS Assay Kit. The integrity of the RNA samples was assessed by BioAnalyzer (Agilent). All RNA integrity numbers (RINs) were greater than 8. The purified RNAs with no visible sign of genomic DNA contamination from the HS Nanochip tracings were used for total RNA library generation.

Furthermore, Illumina TruSeq Stranded Total RNA Sample Prep Kit with Ribo-Zero Gold (catalog no. RS-122-2201) was used to transform RNA into cDNA after removing rRNA and mitochondrial RNA. The RNA-seq libraries were prepared according to the manufacturer’s protocol. Finally, 75 bp paired-end sequencing was performed using the Illumina HiSeq 2500 system.

Gene Abundance Estimate and Differential Gene Expression Analysis.

RNA-seq reads were first mapped to the human genome hg19 using TopHat2 (48). Raw read counts for each gene were quantified by using featureCounts software (49) that uses the GENCODE v.22 Gene Transfer Format (GTF) file as a transcript reference (GENCODE annotation). Genes with counts below 5 for at least two samples out of three within each group were filtered out. Then, the counts were normalized toward the common library size. The count data were assumed to follow a negative binomial distribution. To improve the estimate of overdispersion and to identify genes differentially expressed between samples, R package DESeq2 (50) was used to estimate the smoothed overdispersion parameters and to calculate P values with Wald test for group comparison under a generalized linear model. The P value cutoffs were determined by controlling the mean number of false positives (51).

Western Blotting.

Western blotting was performed according to standard procedures. Antibodies used were MYH9 (Santa Cruz, sc-98978), FOXE1 (Abcam, ab134129), and beta-Actin (Santa Cruz, sc-47778).