Functional characterization of recurrent FOXA2 mutations seen in endometrial cancers

Abstract

FOXA2, a member of the forkhead family of DNA-binding proteins, is frequently mutated in uterine cancers. Most of the mutations observed in uterine cancers are frameshifts and stops. FOXA2 is considered to be a driver gene in uterine cancers, functioning as a haploinsufficient tumor suppressor. The functional consequences of FOXA2 mutations, however, have not yet been determined. We evaluated the effects that frameshift mutations as well as a recurrent missense mutation have on FOXA2 transcriptional activity. Recurrent N-terminal frameshifts resulted in truncated proteins that failed to translocate to the nucleus and have no transcriptional activity using an E-cadherin/luciferase reporter assay. Protein abundance was reduced for the recurrent p.S169W mutation, as was transcriptional activity. A C-terminal frameshift mutation had increased FOXA2 levels evidenced by both Western blot and immunofluorescence. Given that FOXA2 is a recognized activator of E-cadherin (CDH1) expression and E-cadherin’s potential role in epithelial-to-mesenchymal transition in a wide range of cancer types, we tested the hypothesis that FOXA2 mutations in primary uterine cancer specimens would be associated with reduced CDH1 transcript levels. qRT-PCR revealed significantly lower levels of CDH1 expression in primary tumors with FOXA2 mutations. Our findings in vitro and in vivo suggest that reduced transcriptional activity associated with FOXA2 mutations in uterine cancers is likely to contribute to pro-tumorigenic changes in gene expression.

INTRODUCTION

Endometrial cancer (EC) is the most common gynecologic malignancy in the United States.¹ In contrast to many other common cancers, both incidence and mortality are increasing for EC.² A similar increase in EC burden has been described for most developed countries.³ Obesity and metabolic syndrome, as well as our aging population, are key factors driving the growing incidence of EC.^4–6 Exposure to endogenous estrogens (associated with obesity) is an important contributing factor.^{7, 8} The role that estrogens play in the development of EC is further highlighted by the fact that germline variants in the estrogen receptor (ESR1) are associated with risk for EC (most strongly with the endometrioid subtype)⁹ and that somatic ESR1 mutations are seen in primary EC.^{10, 11}

FOXA2 is a member of the FOXA subfamily of forkhead box proteins that includes FOXA1, FOXA2 and FOXA3.^12,13 FOXA proteins bind DNA and act as pioneer and transcription factors.¹⁴ All three forkhead box proteins cooperate with hormone or nuclear receptors to regulate gene activation.^12,15 The interplay between FOXA2 and ERα in regulating gene expression is complex.¹⁶ FOXA1 (HNF3A) and FOXA2 (HNF3B) were initially characterized as essential factors for liver development.¹⁷ Members of the FOXA subfamily have also been shown to play important roles in the development of lung, pancreas, breast and uterine tissues.^18–21 There is growing appreciation of their roles in tumor initiation and progression. Li and colleagues²² demonstrated the important interplay between hormone receptors and FOXA proteins and the implications for different rates of hepatocellular cancer in males and females. FOXA2 influences metastatic potential in primary lung and colon carcinomas.^{23, 24} The subfamily of proteins also play important roles in tumor cell metabolism, including the regulation of lipid uptake in breast cancer cells and effects on cell growth.²⁵ The potential for therapeutic intervention by targeting FOXA proteins is unclear given the complexity of the transcription factor signaling they control and context dependent roles for FOXA protein.¹²

FOXA2 mutations are seen in uterine (endometrial) carcinomas (EC) and carcinosarcomas.^26–28 Based on the pattern and frequency of mutations observed in primary tumors, FOXA2 is considered a driver of endometrial tumorigenesis. The convergence of perturbed Wnt-signaling and FOXA2 function in one of the earliest histologically identifiable precancerous lesions in the endometrium (hyperplasia) can be taken as evidence that FOXA2 functions in the epithelial component of the endometrium and is important in maintaining tissue homeostasis. Beta-catenin (CTNNB1), which is frequently mutated in endometrial carcinomas, has been shown to play a role in the regulation of FOXA2 expression in endometrial hyperplasia, a recognized precursor of adenocarcinoma.²⁹ FOXA2 is central to the earliest stages of development of the endometrium and is a critical mediator in uterine gland differentiation.³⁰ In view of FOXA2’s role in uterine development, maintenance of tissue homeostasis and cooperativity with the estrogen receptor in regulation of gene expression, it is not surprising that disruption of FOXA2 function is seen in EC. We, and others, have previously shown that several FOXA2 mutations are common in primary uterine cancer specimens. We, therefore, undertook studies to assess the effect of these of recurrent endometrial cancer FOXA2 mutations on protein function.

MATERIALS AND METHODS

Cell culture

HEK293T cells were cultured in DMEM (Life Technologies, Carlsbad, CA), HEC-1A cells were cultured in McCoy’s 5A media (Life Technologies, Carlsbad, CA), and Ishikawa cells were cultured in 1:1 F12:DMEM (Life Technologies, Carlsbad, CA). All cell lines were cultured with 10% FBS (Life Technologies, Carlsbad, CA). Cell lines used were confirmed to be mycoplasma negative using the MycoAlert Mycoplasma Detection Kit (Lonza, Basel, Switzerland).

Transfection

Transfections for luciferase reporter assay, immunofluorescence microscopy, and cyclohexamide-chase experiments performed using Lipofectamine® 2000 (Invitrogen, Carlsbad, CA) according to manufacturer’s instructions. Quantity of plasmid transfected noted in figure legend of relevant figures.

Luciferase reporter assays

Cells were seeded and transfected with empty vector, WT and mutant FOXA2 expression plasmids (1 microgram), pRL-CMV renilla luciferase control (20ng, Promega, Madison, WI), and the pGL3 E-cadherin luciferase construct (1 microgram). Luciferase was measured via the Dual-Luciferase 1000 System (Promega, Madison, WI).

Immunofluorescence microscopy

Transfected cells plated on coverslips were fixed and permeabilized using standard methods. Exogenous FOXA2 was detected using the 9E10 anti-myc antibody. DAPI was used to stain nuclei. Imaging was performed using a Keyence BZ-9000 florescent microscope (Keyence Corporation, Itasca, IL, USA). An Olympus FV 1000 Filter Confocal system (Olympus America Inc, Melville, NY, USA) was used for confocal imaging.

qRT-PCR

RNA prepared from primary tumor samples was reverse transcribed (Supplemental Table 1) and SYBR Green qRT-PCR assays were performed to assess CDH1 and TBP (control) transcript levels (Bio-Rad CFX96 Real-Time System). Relative CDH1expression level was calculated using the ΔΔCt method.

RESULTS

Previous studies have indicated that FOXA2 is a probable pathogenic driver in uterine cancer. The majority of FOXA2 mutations are loss-of-function defects (frameshift and nonsense alleles), consistent with FOXA2’s role as a tumor suppressor.^{26, 27} Primary tumor specimens rarely have a second mutation or other identifiable abnormality. Haploinsufficiency for FOXA2 could be sufficient to drive endometrial tumorigenesis, as has been proposed for a number of tumor suppressors.^{31, 32} We focused our analysis of mutant FOXA2 proteins on highly recurrent mutations.^26–28

Recurrent FOXA2 mutations have variable effects on protein localization and stability

Myc-tagged FOXA2 expression constructs were developed for four recurrent mutations that together are representative of the most frequently seen endometrioid endometrial cancer mutations²⁷ (Figure 1a). Two frameshifts N-terminal of the forkhead domain (both involving p.Q122) and the recurrent p.S169W missense allele in the forkhead/DNA binding domain were studied. The A353Pfs*8 mutation was chosen as a representative of a cluster of insertions/deletions involving coding sequences c.1053–1075 (NM_021784) that result in frameshifts and loss of the C-terminal transcriptional activation domain. Western blot analysis of HEK293T cells transiently transfected with wild-type (WT) and the four FOXA2 mutant expression constructs revealed that expression levels of exogenous FOXA2 were variable. The 3X Myc-tagged A353Pfs*8 protein was strikingly more abundant, and the short Q122Rfs*15 was scarcely detectable (Figure 1b).

Figure 1: Recurrent FOXA2 mutations effect protein localization and abundance.

(a) Location of mutants investigated relative to the predicted nuclear localization signal, transactivation and forkhead domains. NLS: nuclear localization signal, TAD (transactivation domain), FHD (forkhead domain). (b) Western blot analysis of Myc-tagged wild-type and mutant FOXA2 in transiently transfected HEK293T cells. Approximate sizes of the exogenous FOXA2 proteins are given. Vinculin used as loading control (c) Immunofluorescence (anti-Myc antibody) showing nuclear localization of wild-type, p.S169W and p.A353Pfs*8 and cytoplasmic localization of p.Q122Afs*123 and p.Q122Rfs*15 FOXA2. Scale bar: 40 µm. (d) Confocal images showing Myc-tagged FOXA2 localization. Scale bar: 10µm.

A FOXA2 wild-type expression construct (pcDNA3 with a N-terminal 3xmyc tag) was generated using HepG2 RNA and RT-PCR (SuperScript III First-Strand Synthesis Invitrogen, Carlsbad, CA, USA using primers for the long FOXA2 isoform (NM_021784) F:TAAGCAGAATTCGCCATGCACTCGGCTTCCAGT and R: TAAGCAGATATCCCTGAAGCCGTCGTCTTCTT). Mutants were generated using QuikChange II XL-Site Directed Mutagenesis (Agilent Technologies, Santa Clara, CA, USA). The mutagenesis primers were as follows: (S169W) F:ATGAGCGAGATGTACCAGTAGGGCGGCTTG and R:CAAAGCCGCCCTACTGGTACATCTCGCTCAT; (A353Pfs*8) F:CAGCAGGCCGCCCCACCACCCG and R:CGGGTGGTGGGGCGGCCTGCTG; (Q122delG) F:CGCTCGGGGGCAGGCGGCCG R:CGGCCGCCTGCCCCCGAGCG; (Q122insG) F:GCTCGGGGGGGCAGGCGGCCG and R:CGGCCGCCTGCCCCCCCGAGC).

HEK293T cells were plated in 6-well tissue culture plates and transfected with 500ng of each plasmid DNA. Protein lysates for Western blot analysis were prepared 36hr later. Exogenous FOXA2 was detected with anti-myc antibody (anti-c-myc, 1:3000, sc-40, Santa Cruz Biotechnology, Dallas, TX, USA). Vinculin was used as a loading control. Blots were imaged using an Odyssey Fc Imaging System (LICOR Biosciences, Lincoln, NE, USA).

For immunofluorescence HEK293T cells were plated onto 6-well plates at equal confluence on coverslips. Cells were transfected with 500 ng of each plasmid DNA and allowed to grow for 24 hours. Cells were fixed with 3.7% formaldehyde, permeabilized with 0.1% Triton-X-100 for 10 minutes, and blocked with 4% normal goat serum for 10 minutes. Primary antibody for c-myc tag (1:50, sc-40, Santa Cruz Biotechnology, Dallas, TX, USA) was incubated 1 hour at room temperature (RT). Secondary antibody, Alexa Fluor 488 goat anti-mouse (1:250, Invitrogen, Carlsbad, CA, USA) was incubated in the dark at RT for 30 minutes. Nuclei were stained with DAPI. Slides were visualized using Keyence BZ-9000 florescent microscope (Keyence Corporation, Itasca, IL, USA). Each image (40X) is a representative of 5 fields. The confocal images were generated using an Olympus FV 1000 Filter Confocal system (Olympus America Inc, Melville, NY, USA).

Immunofluorescence (IF) confirmed that the Myc-tagged WT protein localized to the nucleus (Figure 1c & d). The N-terminal frameshift mutants lacking the predicted nuclear localization signal (amino acids 254–264) were primarily cytoplasmic, with little or no nuclear signal detectable (right panels Figure 1c & d). The A353Pfs*8 and the S169W proteins, on the other hand, were localized to the nucleus. Studies in the Ishikawa endometrial cancer cell line confirmed the HEK293T findings for localization of exogenous FOXA2 proteins (Supplemental Figure 1) and differences in abundance (Supplemental Figure 2) as seen in HEK293T.

High level expression of p.A353Pfs*8 is not associated with increased protein half-life

The increased abundance of FOXA2 A353Pfs*8 mutant protein (Figure 1b, Supplemental Figure 2) led us to assess the half-life of the exogenous FOXA2 proteins (WT, S169W and A353Pfs*8, of which all are translocated to the nuclear compartment and could thus retain transcriptional activity). Cycloheximide treatments beginning 24 hours post-transfection of Ishikawa cells, coupled with Western blot analysis of the exogenous FOXA2 proteins at 0, 1, 2, 4 and 8 hours post cycloheximide treatment, revealed that both the p.A353Pfs*8 and pS169W proteins had slightly reduced half-lives relative to wild-type FOXA2 (Figure 2). The half-life of wild-type Myc-tagged FOXA2 was approximately 8 hours, whereas both the A353Pfs*8 and S169W protein had half-lives between 5–6 hours.

Figure 2: Mutant FOXA2 proteins have reduced half-lives.

(a) Representative Western blots for transiently transfected Ishikawa cell samples collected at set time points as noted. The relative stability of FOXA2 WT protein is evident across time points. Beta-actin and cyclin D1 served as controls (negative and positive) respectfully. Inset images for the 8h protein collection from A353Pfs*8 (red outline) indicates that data are from a different blot. The specimen was run on a separate gel at the same time as the 0–4 h preparations. Reductions in Cyclin D1 levels are consistent with findings from previous studies. (b) Quantitation of FOXA2 levels for three independent experiments suggests both the S169W and A353Pfs*8 protein have reduced half-levels relative to exogenous FOXA2.

Ishikawa cells were plated in 12-well cell culture plates at equal confluency. Cells were transfected with 1µg of each plasmid (wild-type, S169W, and A353Pfs*8). Twenty-four hours later 50µg of cycloheximide (Calbiochem, EMD Millipore, Billerica, MA, USA) was added. Cells were trypsined and lysates prepared at the specificed times. Proteins were size separated on 10% SDS-PAGE gels. 30µg of protein were loaded for wild-type and the A353Pfs*8 mutant, 40µg of S169W protein was loaded to ensure sufficient exogenous FOXA2 was present for the time course. Primary antibodies used (and concentrations) are as follows: c-myc antibody, SC-40 (1:1000) [Santa Cruz Biotechnology, Dallas, TX, USA]; anti-beta-actin (1:500); and anti-cyclin D (1:100) [Abcam, Cambridge, MA, USA]. Secondary antibodies were at 1:12000 for anti-rabbit and anti-mouse from LiCor (Lincoln, NE, USA). Blots were imaged using the LiCor Clx (Lincoln, NE, USA) imager at 700 and 800 wavelengths. Protein quantification performed using the LiCor Image Studio (Lincoln, NE, USA) software. Data plotted represent three independent experiments. SD shown for wild-type experiments.

Recurrent FOXA2 mutations have variable effects on transcription factor activity

FOXA2 mutants that fail to localize to the nucleus and/or that lack the putative transcriptional activation domains would be expected to have impaired transcriptional activity. The frequent loss-of-function mutations seen in EC are likely to reflect selection for multiple pro-tumorigenic changes in gene expression directly or indirectly related to FOXA2 activity. FOXA2 deficiency has been implicated in epithelial-to-mesenchymal transition in a variety of malignancies^33–36 in part due to the direct role of FOXA2 in activation of E-cadherin (CDH1) transcription.^{33, 34}

We used a luciferase reporter assay to test the transcriptional activity of the FOXA2 mutants. The luciferase reporter consists of a 762bp fragment of the CDH1 promoter that includes 4 HNF3 (FOXA) binding sites (Figure 3a). We show that exogenous WT FOXA2 induces luciferase expression by ~2.5-fold compared to EV in Ishikawa endometrial cancer cells (Figure 3b). The N-terminal frameshift mutant FOXA2 proteins (p.Q122Afs*123 and p.Q122Rfs*15) had no increase in luciferase activity similar to EV, consistent with evidence that these mutations are excluded from the nucleus (Fig 3B). The p.S169W mutant FOXA2 was associated with a modest but statistically significant (P<0.05) reduction in luciferase activity (Figure 3b). The C-terminal frameshift p.A353Pfs*8 mutant protein, however, was associated with increased luciferase activity (Figure 3b). The specificity of the luciferase reporter for FOXA2 binding was confirmed by deletion of the FOXA2 binding sites, abrogating WT FOXA2 induction of luciferase expression in Ishikawa, RL-952 and HEC1A endometrial cell lines (data not shown).

Figure 3: Mutant FOXA2 proteins are associated with reduced E-cadherin expression.

(a) Schematic of luciferase reporter assay using a 762bp genomic fragment from the E-cadherin (CDH1) promoter (adapted from Liu et al,⁵⁰). The location of the four FOXA binding sites (green boxes) and FOXA2 transcription start site (TSS) are shown. (b) Luciferase reporter assay demonstrates significant reduction in luciferase activity for the p.S169W and pQ122 frameshift mutations relative to wild-type (WT) FOXA2. The A353Pfs*8 on the other hand shows increased luciferase activity. Results for triplicate transfections of Ishikawa (luciferase/renilla ratios) normalized to the empty vector (EV) are presented with standard deviations. * P <.05, two-tailed t-test with Welch’s correction. (c) Reduced CDH1 expression in primary endometrial cancers with FOXA2 mutations. qRT-PCR for 22 wild-type endometrioid endometrial cancers and 14 with FOXA2 mutant tumors. Assays were performed in triplicate and expression normalized to TBP. * P <.001, two-tailed t-test with Welch’s correction.

Luciferase reporter assays were performed in the Ishikawa cell line. Cells were co-transfected with a pcDNA empty vector, FOXA2 wild-type plasmid, or one of four mutant FOXA2 plasmids [1µg] along with pRL_CMV [20ng] (Promega, Madison, WI, USA) and pGL3 promoter E-cadherin luciferase [1µg].⁵⁰ 24h following transfection, cells were lysed, the lysates cleared by centrifugation and 20µL of supernatant transferred to 96-well plates. Activity was measured using the Dual-Luciferase Reporter 1000 assay system (Promega, Madison, WI, USA) and Veritas Microplate Luminometer (Turner BioSystems, Sunnyvale, CA, USA). All data represent luciferase/renilla ratios normalized to the EV control. Experiments were performed in triplicates and data plotted showing SD.

qRT PCR were performed in triplicate and normalized to TBP. Analysis performed using Bio-Rad CFX Manager v3.1 (Hercules, CA, USA).

To determine if FOXA2 mutation in primary endometrial cancers is associated with altered CDH1expression, we assessed transcript levels in endometrioid tumors, 14 of which carry FOXA2 mutations and 22 WT tumors. Both groups of tumors (mutant and WT) were from women who primarily had early stage disease (I or II). Most of the FOXA2 mutant tumors were grade 2, whereas grade 1 tumors were most common in the wild-type group. The majority of FOXA2 mutations were frameshifts (12 of 14 investigated) with one stop (p.Q193*) and one missense mutation (p.E367D) (Supplemental Table 1). qRT-PCR revealed that CDH1 transcript levels were significantly lower in mutant tumors than wild-type tumors (P<.001, Figure 3c). Among the mutant tumors, the specimen with the p.A353Pfs*8 mutation had the highest expression of CDH1, consistent with increased luciferase reporter assay in vitro. It is noteworthy that the single tumor with two FOXA2 mutations (p.M142Nfs*103 and p.Q318Pfs*50) had the lowest CDH1 expression.

DISCUSSION

The preponderance of loss-of-function FOXA2 mutations seen in EC, coupled with the observation that mutations or epigenetic changes involving the second allele are infrequent events in primary tumors, is consistent with FOXA2’s role as a haploinsufficient tumor suppressor. FOXA2 mutations in EC occur in the last exon (2 coding exons) and thus transcripts from the stop/frameshift mutations would not be subject to nonsense-mediated decay.^{26, 27, 37} Changes in abundance and activity of the truncated/frameshift mutant proteins are expected given loss of C-terminal amino acids and/or addition of novel sequences in the frameshift proteins. Previous studies have shown that FOXA2 stability is determined by both sumoylation and acetylation, and is context dependent.^{38, 39} The variable expression of exogenous mutant FOXA2 protein that we observed in endometrial cell lines (Figure 1b, Supplemental Figure 2) could reflect changes in translational efficiency, protein stability, and potentially differences in transcript stability. It is noteworthy that earlier work from our group demonstrated that FOXA2 transcript levels were higher in primary ECs with mutations than WT tumors.²⁷ Transcript abundance, however, did not correlate with protein levels. Increased FOXA2 transcription (higher transcript levels) in tumors with mutations could reflect a feedback mechanism to regulate FOXA2 activity.

The N-terminal frameshift proteins (p.Q122Rfs*15 and p.Q122Afs*123) fail to localize to the nucleus and have no transcriptional activity in a luciferase reporter assay (Figure 1c & d, Figure 3b). Because these mutants lack the DNA binding/forkhead domain and do not translocate to the nucleus, loss of transcriptional activity is not surprising.⁴⁰ Together, these data provide evidence that these recurrent mutations are loss-of-function. They do not, however, provide functional data to support a model of FOXA2 haploinsufficient tumor suppression.

We focused our efforts to determine the effects of mutations on FOXA2’s transcriptional activity on the CDH1 gene. E-cadherin (CDH1) plays key roles in cell-cell adhesion and tissue organization.⁴¹ Loss of cell-cell adhesion associated with CDH1 abnormalities contributes to both the development and progression of cancers.⁴² FOXA2 is a positive regulator of CDH1 expression and altered FOXA2 activity has been linked to epithelial-to-mesenchymal transition (EMT) in breast and pancreatic cancer.^{33, 34} It is important to note, however, that it remains uncertain if loss of CDH1 drives EMT or if dysregulation of CDH1 is a result of EMT.⁴³

Findings from our luciferase reporter assays (Figure 3b) and analysis of CDH1 in primary endometrial cancers (Figure 3c) demonstrate that EC FOXA2 mutations result in proteins with reduced transcription factor activity that is correlated with target gene expression in primary samples. The missense mutant we studied (p.S169W within the DNA binding domain) had significantly reduced in vitro transcriptional activity relative to WT FOXA2. Serine 169 is potential phosphorylation site²⁷ and it is possible that phosphorylation at this site increases transcriptional activity or influences protein stability. Given the fact that the level of expression of the p.S169W mutant protein was much lower than that for WT FOXA2 in transfection assays (Figure 1b), it is most likely that the relative abundance of p.S169W FOXA2 explains the reduced transcriptional activity seen with the CDH1/luciferase reporter assay. The significant increase in luciferase activity with the A353Pfs*8 mutant, in which the first transcriptional activation domains remains intact (see Figure 1a), may reflect the increase in expression of the mutant protein (Figure 1b) relative to WT FOXA2. It is noteworthy that among the primary EC specimens with FOXA2 mutations, tumor 1929T carrying the p.A353Pfs*8 mutation had the highest expression of CDH1 (Figure 3c and Supplemental Table 1). On the other hand, the single tumor in our study that harbored two FOXA2 mutations (O-1032T) had the lowest level of CDH1 expression. These cases provide support for a model in which FOXA2 activity in primary tumors influences CDH1 transcriptional activity. The observed reduction in CDH1 expression in tumors heterozygous for FOXA2 mutation provides evidence for haploinsufficiency with respect to transcriptional activity. We recognize that CDH1 transcript levels in EC specimens lacking FOXA2 mutations vary extensively (most ECs express FOXA2²⁷), and that many factors in addition to FOXA2 regulate CDH1 expression. Our discovery that FOXA2 mutations are associated with lower CDH1 expression in primary EC specimens is, to the best of our knowledge, the first demonstration of a potential selective advantage for FOXA2 mutations in EC. Reduced CDH1 expression in a wide range of tumor types, including EC, has been associated with variety of negative prognostic features and poor outcomes. ^44–49 Tumor tissue sections were not available to evaluate the primary tumors with FOXA2 mutations for changes in CDH1 protein levels or evidence of EMT. Nonetheless, our data on CDH1 expression in vitro and in vivo are consistent with loss of FOXA2 transcriptional activity associated with EC mutations. Additional studies will be required to understand the intersection between FOXA2 mutation and altered pioneer and transcription factor activity in EC.

Novelty and Impact:

Somatic mutations in FOXA2 in uterine cancers are potential “driver” events, but the functional consequences of those mutations have not been elucidated. We determined that FOXA2 mutations, representative of those most frequently seen in uterine cancers, result in altered protein abundance and transcription factor activity. Importantly, we show that in primary tumors with FOXA2 mutations, expression of the known transcriptional target E-cadherin is reduced and provide evidence that FOXA2 is a haploinsufficient tumor suppressor.

Abbreviations:

EC

endometrial cancer

WT

wild-type

CDH1

E-cadherin

TBP

TATA-binding protein

FOXA2

Forkhead box A2

qRT-PCR

quantitative reverse transcription – polymerase chain reaction