Translation of Human Tissue Factor Pathway Inhibitor-β mRNA Is Controlled by Alternative Splicing Within the 5′ Untranslated Region

Abstract

Objective

Tissue factor pathway inhibitor (TFPI) blocks the initiation of coagulation by inhibiting TF-activated factor VII, activated factor X, and early prothrombinase. Humans produce two 3′ splice variants, TFPIα and TFPIβ, which are differentially expressed in endothelial cells and platelets and possess distinct structural features affecting their inhibitory function. TFPI also undergoes alternative splicing of exon 2 within its 5′ untranslated region. The role of exon 2 splicing in translational regulation of human TFPI isoform expression is investigated.

Approach and Results

Exon 2 splicing occurs in TFPIα and TFPIβ transcripts. Human tissue mRNA analysis uncovered a wide variability of exon 2 expression. Polysome analysis revealed a repressive effect of exon 2 on TFPIβ translation but not on TFPIα. Luciferase reporter assays further exposed strong translational repression of TFPIβ (90%) but not TFPIα. Use of a Morpholino to remove exon 2 from TFPI mRNA increased cell surface expression of endogenous TFPIβ. Exon 2 also repressed luciferase production (80% to 90%) when paired with the β-actin 3′ untranslated region, suggesting that it is a general translational negative element whose effects are overcome by the TFPIα 3′ untranslated region.

Conclusions

Exon 2 is a molecular switch that prevents translation of TFPIβ. This is the first demonstration of a 5′ untranslated region alternative splicing event that alters translation of isoforms produced via independent 3′ splicing events within the same gene. Therefore, it represents a previously unrecognized mechanism for translational control of protein expression. Differential expression of exon 2 denotes a mechanism to provide temporal and tissue-specific regulation of TFPIβ-mediated anticoagulant activity.

Tissue factor pathway inhibitor (TFPI) is a trivalent Kunitz-type serine protease inhibitor. It is the major regulator of tissue factor (TF)–induced blood clotting, inhibiting activated factor VII within the TF-activated factor VII complex via its first Kunitz domain (K1) and activated factor X via its second Kunitz domain (K2).¹ In addition, it has recently been shown to rapidly inhibit forms of prothrombinase generated in the early stages of coagulation.^2,3 These interactions serve to limit downstream thrombin generation and fibrin formation to the site of vascular injury, thereby minimizing the development of disseminated intravascular coagulation and the formation of occlusive thrombi. The physiological importance of TFPI is highlighted in mice lacking K1, which die in utero because of yolk sac hemorrhage and, presumably, consumptive coagulopathy.⁴ Furthermore, tissue-specific knockouts of K1 demonstrate that both endothelial⁵ and hematopoietic cell^6,7 (presumably platelet) TFPI are important regulators of thrombus formation in vivo.

Alternative splicing at the 3′ end of the TFPI gene gives rise to 2 TFPI isoforms, TFPIα and TFPIβ,⁸ which differ only at their C termini. The C-terminal region of TFPIα contains a third Kunitz domain (K3) followed by a stretch of highly basic amino acids that bind to an acidic region of the factor V B-domain, providing a key exosite interaction required for inhibition of early forms of prothrombinase.^2,3 The C terminus of TFPIβ encodes a glycosylphosphatidylinositol anchor that facilitates direct binding to the cell surface.⁸ Compared with TFPIα, TFPIβ is an equally effective inhibitor of TF-dependent procoagulant activity but is a significantly more potent inhibitor of TF-dependent cellular migration, suggesting that it may function in the regulation of TF-mediated cellular signaling events.⁹

Translational control of protein expression is important in many physiological processes, including temporal and spatial protein expression, response to cellular stimuli, embryonic development, and disease.¹⁰ It is predominantly mediated by regulatory elements within the 5′ or 3′ untranslated regions (UTRs) of the target mRNA.¹¹ Eukaryotic 5′ UTRs have an average length of 20 to 100 nucleotides.¹² Longer 5′ UTRs allow for more efficient translation initiation because of preloading of 43S ribosomal initiation complexes.¹³ However, they may also be poorly translated because of the inclusion of negative regulatory elements.^14,15 Alternative splicing within the 5′ UTR occurs in ≈12% of human mRNA species,¹⁶ allowing for the inclusion or removal of translational regulatory elements similar to the inclusion or removal of microRNA-binding sites within 3′ UTRs via alternative polyadenylation.^17,18

The human TFPI gene is ≈85 kb in size and contains 9 introns and 10 exons.¹⁹ Exons 1 and 2 consist of 293 and 122 nucleotides, respectively, and encode the 5′ UTR. Interestingly, in some transcripts, exon 2 is removed via alternative splicing,²⁰ although the relevance of this splicing event has not been previously investigated. Exons 3 through 7 encode the regions common to TFPIα and TFPIβ. Exon 8 encodes the region specific to TFPIβ, whereas exons 9 and 10 encode the regions specific to TFPIα. Two TFPIα mRNA species, 4.0 and 1.4 kb in size, are produced via alternative polyadenylation²⁰ and differ only in the length of their 3′ UTR (2.7 kb and 86 bp, respectively). The TFPIβ message size has not been previously reported.

TFPIα and TFPIβ mRNAs are produced in a 10:1 ratio in adult human and mouse tissues.²¹ However, adult mice produce only TFPIβ protein in all major vascular beds,²² with TFPIα found only in platelets.²³ These data suggest that TFPI isoform production in mice is translationally controlled. The predominant human isoform is less clear. Several studies suggest that it is TFPIα because it is found in plasma,²⁴ platelets,²³ and placenta²⁵ and heparin-releasable TFPI is TFPIα.²⁶ However, human endothelial cell lines produce TFPIβ,²⁷ suggesting that TFPIβ is the predominant endothelial isoform. Furthermore, there is evidence suggesting that TFPI production is increased via translational mechanisms in response to serum,²⁸ basic fibroblast growth factor/heparin,²⁹ and adiponectin.³⁰ The current studies were designed to further investigate the role of exon 2 splicing in translational control of human TFPI isoform expression and how it relates to the distinct tissue-specific expression patterns of TFPIα and TFPIβ.

Results

Exon 2 Is Present in TFPIα and TFPIβ Message

The human TFPI 5′ UTR consists of 2 exons. Northern blot analysis of human lung RNA was performed to determine whether exon 2 (122 bp) is present exclusively in TFPIα or TFPIβ message using probes directed toward exon 1, exon 2, exon 6 (total TFPI), exon 8 (TFPIβ), and exon 9 (TFPIα; Figure 1A). Two bands, 1.4 and 4.1 kb, were detected using the exon 6 (total TFPI) probe. These bands, also identified by the exon 9 (TFPIα) probe, arise from alternative polyadenylation within the TFPIα 3′ UTR.²⁰ These 2 bands were also detected by the exon 1 and exon 2 probes, demonstrating that exon 2 is present in TFPIα message (Figure 1B).

Figure 1.

Northern blot and nested polymerase chain reaction (PCR) analyses demonstrate that exon 2 is present in tissue factor pathway inhibitor-α (TFPIα) and TFPIβ mRNA. A, Schematic of TFPIα and TFPIβ exon structure. Black lines represent the location of probes used for Northern analysis. Black and dark gray arrows represent primers used to amplify TFPIα and TFPIβ, respectively, in the first round of nested PCR. Light gray arrows represent primers used to amplify the region spanning exons 1 through 3 in the second round of nested PCR. Product sizes produced in each round of nested PCR are indicated above both schematic diagrams. *AUG start codons; #in-frame TGA stop codons. B, Northern blot for TFPI transcript using the indicated probes. The 1.4- and 4.1-kb bands represent the 2 major TFPIα mRNA species that arise from alternative polyadenylation within the TFPIα 3′ untranslated region.²⁰ The 1.1-kb band represents TFPIβ message. To adequately visualize the less abundant TFPIβ message, the blot probed with the exon 8 probe was exposed to film for an additional week. The lower gel image demonstrates equal RNA loading between lanes and excellent RNA integrity. C, Nested PCR to detect exon 2 splicing in TFPIα and TFPIβ using primers depicted in A. The lower band (293 bp) is exon 1 spliced to exon 3, the middle band (344 bp) exon 1 spliced to exon 3 amplified by residual forward primer from the first round of PCR amplification, and the upper band (415 bp) exon 1 spliced to exon 2 spliced to exon 3. Repeated attempts to identify the uppermost band in the TFPIβ lane were not successful. However, it is likely exon 1 spliced to exon 2 spliced to exon 3 that was amplified by residual outside exon 1 primer, as occurred to produce the 344-bp band.

The exon 8 (TFPIβ) probe identified a 1.1-kb band (Figure 1B), which correlates with the size predicted from its GenBank sequence (NM_001032281.2). However, this band was not recognized by other probes, including those targeting exon 1 or exon 6 (total TFPI), which are present in TFPIβ mRNA. It is possible that probes other than the exon 8 (TFPIβ) probe were not sensitive enough to detect the comparatively lesser amount of TFPIβ message present in human tissue or the 1.1-kb TFPIβ band could not be differentiated from the more intense 1.4-kb TFPIα band that is also recognized by the exon 1 and exon 6 (total TFPI) probes. Therefore, nested polymerase chain reaction (PCR) was performed to determine the presence of exon 2 in TFPIα and TFPIβ transcripts (Figure 1C). In these experiments, a region spanning the start of exon 1 to either exon 8 (TFPIβ) or exon 9 (TFPIα) was first amplified, and then the middle of exon 1 to exon 3 was amplified (Figure 1A). Sequence analysis of the 3 bands (293, 344, and 415 bp) amplified from TFPIα and TFPIβ revealed that exon 2 is present in the 415-bp band from both TFPI isoforms.

Tissue Distribution of TFPIα and TFPIβ Transcripts

The presence or absence of exon 2 in TFPIα and TFPIβ mRNA in 15 human tissues was examined using the first round of nested TFPIα and TFPIβ PCR (Figures 1A, 2A, and 2C). This PCR produces 2 bands for each isoform, corresponding to those containing or lacking exon 2. Because both bands are amplified using the same PCR primer set, comparison of the intensity of the resulting products is an acceptable method to determine the relative amount of each product present.^21,31 Densitometry of the resulting bands revealed that both TFPIα and TFPIβ mRNAs contain a relative paucity of exon 2 in all tissues (Figure 2B and 2D). The amount of TFPIα message lacking exon 2, relative to that containing exon 2, was relatively consistent between adult tissues (Figure 2B), with an average ratio of ≈1.8:1. The presence of exon 2 in TFPIβ mRNA between the tissues was more variable. Testis, thymus, and brain had the highest relative amount of exon 2–containing message, whereas fetal brain and liver and adult heart, kidney, and liver had the least.

Figure 2.

Tissue distribution of tissue factor pathway inhibitor-α (TFPIα) and TFPIβ mRNA containing and lacking exon 2. A, Exon 2 expression in TFPIα mRNA in the indicated tissues, determined using the exon 1 to TFPIα polymerase chain reaction (PCR; Figure 1C). B, Densitometry analysis of A, expressed as a ratio of TFPIα mRNA lacking exon 2 to TFPIα mRNA containing exon 2. C, Exon 2 expression in TFPIβ mRNA in the indicated tissues, determined using the exon 1 to TFPIβ PCR (Figure 1C). D, Densitometry analysis of C,expressed as a ratio of TFPIβ mRNA lacking exon 2 to TFPIβ mRNA containing exon 2. For B and D, the break in the graph represents a ratio of 1. Gel images and densitometry are representative of 2 separate experiments. Fe indicates fetal; Sk, skeletal; and Sp, spinal.

Exon 2 Represses Translation of TFPIβ But Not TFPIα

Polysome analysis was performed on human umbilical vein endothelial cell lysates to examine the translational efficiency of TFPIα and TFPIβ mRNA. Human umbilical vein endothelial cells were used because they are nontransformed cells, express TFPI, and have been used previously for studies of TFPI in endothelial cells.^27,32–35 Total cellular RNA separated into 2 peaks, representing polysomes and 80S ribosomal RNA, after centrifugation on a 20% to 60% sucrose gradient (Figure 3A, black line). In control experiments, human umbilical vein endothelial cell lysates were pretreated with EDTA to disrupt ribosome binding, resulting in the expected shift of all RNA from polysomes to the top of the gradient (Figure 3A, light gray line). RNA was isolated from each fraction and analyzed for the distribution of β-actin (nontranslationally regulated)³⁶ and transforming growth factor β1 (translationally repressed)³⁷ transcript to examine polysome integrity. β-actin mRNA was evenly distributed across fractions 4 to 18 of the gradient (Figure 3B; Figure 3C and 3D, black dashed line), confirming that it is not translationally regulated. The majority of transforming growth factor β1 was in the ribosome-free RNA fractions (16 and 17), with a small amount in translationally inefficient light polysomes (fractions 9 through 11; Figure 3B; Figure 3C and 3D, gray dashed line), confirming that it is translationally repressed. As expected, EDTA pretreatment resulted in a shift of β-actin and transforming growth factor β1 mRNA to the ribosome-free RNA region (Figure 3B).

Figure 3.

Polysome analysis demonstrating that tissue factor pathway inhibitor-β (TFPIβ), but not TFPIα, is translationally repressed by exon 2. A, Polysomes were isolated from human umbilical vein endothelial cell lysates and RNA measured in each fraction. Lysates were untreated (black) or pretreated with EDTA (gray) to disrupt ribosome RNA binding. B, Polymerase chain reaction analyses of fractions to detect β-actin, transforming growth factor (TGF) β1, TFPIα, and TFPIβ. The lower band in TFPIα and TFPIβ gel images is mRNA lacking exon 2, whereas the upper band is mRNA containing exon 2. For the EDTA-treated (+EDTA) sample, only fractions ≥10 were analyzed because the lower fractions (1–9) did not contain detectable RNA (A₂₆₀ <0.01). Band intensity from B was quantified by densitometry and the amount of TFPIα (C) or TFPIβ (D) mRNA containing or lacking exon 2 in each fraction expressed as a percentage of the total amount of the same product across the entire gradient. Identical β-actin and TGF-β1 controls are shown on both graphs.

To examine the translational efficiency of TFPIα and TFPIβ message, polysome fractions were analyzed by PCR with conditions used in the first round of nested TFPIα and TFPIβ PCR described above (Figure 1A). TFPIα message containing (1139 bp; Figure 3B and 3C, solid gray line) and lacking (1017 bp; Figure 3B and 3C, solid black line) exon 2 was present in heavy polysomes, with message containing exon 2 shifted slightly up the gradient (initially observed in fraction 6 versus fraction 5; Figure 3B and 3C). This suggests that exon 2 may have a minor, if any, repressive effect on TFPIα translational efficiency. TFPIβ transcript lacking exon 2 (1001 bp; Figure 3B) was predominantly present in heavy polysomes (initially observed in fraction 4; Figure 3B and 3D, solid black line). In contrast, TFPIβ message containing exon 2 (1123 bp; Figure 3B) was distinctly shifted into light polysomes (initially observed in fraction 8; Figure 3B and 3D, solid light gray line) where RNA is translated relatively inefficiently, suggesting that exon 2 is a negative translational regulatory element when present in TFPIβ message.

Exon 2 Is a Strong Translational Repressor When Coupled With the TFPIβ 3′ UTR

A luciferase reporter system was established to quantify exon 2–mediated translational repression of TFPIα and TFPIβ protein production. The pCMV-Gluc vector was modified to express green fluorescent protein (Figure III in the online-only Data Supplement) to normalize and compare luciferase activity between cell lines. Initially, 6 constructs were produced: 3 had the long TFPIα 3′ UTR, corresponding to that present in the 4.1-kb message, and 3 had the TFPIβ 3′ UTR, inserted immediately after the luciferase coding sequence. Each 3′ UTR was paired with either no corresponding 5′ UTR or a 5′ UTR consisting of exon 1 (Ex1) only or exons 1 and 2 (Ex1+Ex2; Figure 4A). Constructs were stably transfected into chinese hamster ovary (CHO) cells and luciferase activity measured to assess the effect of 5′ and 3′ UTR combinations on luciferase production. Constructs containing only the long TFPIα or TFPIβ 3′ UTR (no 5′ UTR) had near-identical luciferase activity (Figure 4B), suggesting that the 3′ UTR alone is not responsible for translational regulation of either isoform. Constructs containing Ex1 only or Ex1+Ex2 had similar luciferase activity when paired with the long TFPIα 3′ UTR (Figure 4C), confirming polysome results demonstrating that exon 2 has little effect on TFPIα production. However, when paired with the TFPIβ 3′ UTR, the construct containing Ex1+Ex2 had markedly decreased (93%) luciferase activity compared with the corresponding Ex1 construct (Figure 4D), consistent with polysome results demonstrating that exon 2 is a strong negative regulator of TFPIβ protein production.

Figure 4.

Luciferase assays demonstrate that exon 2 represses synthesis of tissue factor pathway inhibitor-β (TFPIβ) but not TFPIα. A, Schematic of 6 luciferase constructs used for chinese hamster ovary (CHO) and EA.hy926 transfections. B, The TFPIα Long or TFPIβ 3′ untranslated region (UTR) alone does not affect luciferase production (P>0.05; n=6). No 5′ UTR is included in these constructs. C and D, Exon 2 of the TFPI 5′ UTR negatively regulates TFPIβ production. No significant difference in luciferase activity was observed between the Ex1-TFPIα Long and Ex1+X2-TFPIα Long CHO cells (C; P>0.05; n=6), whereas a significant decrease was observed using Ex1+X2-TFPIβ CHO cells compared with Ex1-TFPIβ cells (D; P<0.0001; n=6). E and F, Similar results were obtained when the same constructs were transfected into EA.hy926 cells (P>0.05 for Ex1-TFPIα Long vs Ex1+X2-TFPIα Long, n=6; P<0.001 for Ex1-TFPIβ vs Ex1+X2-TFPIβ, n=6). For B through F, luciferase activity was determined as outlined in the Methods section, the average luciferase activity of all control construct replicates for that figure (the left-hand bar of each graph) designated as 100%, and individual luciferase activities compared with this value to give normalized luciferase activity as a percentage. Bars represent mean±1 SD. Ex1 indicates TFPI exon 1; and Ex1+Ex2, TFPI exon 1 spliced to exon 2.

TFPI is produced primarily by endothelial cells in vivo.³⁸ Therefore, the 6 luciferase constructs (Figure 4A) were stably transfected into the human endothelial-like EA.hy926 cell line. In these experiments, luciferase activity was normalized to luciferase mRNA because of low green fluorescent protein expression by the EA.hy926 cells. The results obtained were similar to those of CHO cells. Ex1 only and Ex1+Ex2 produced similar amounts of luciferase activity when paired with the long TFPIα 3′ UTR (Figure 4E). There was a marked decrease in luciferase production (86%) by the Ex1+Ex2 construct compared with the Ex1 construct when paired with the TFPIβ 3′ UTR (Figure 4F), demonstrating that exon 2 is a strong negative regulator of TFPIβ protein production in EA.hy926 cells.

Exon 2 Is a Repressor of Endogenous TFPIβ Expression

To demonstrate that exon 2 is a negative regulator of endogenous TFPIβ but not TFPIα protein production, a Morpholino targeting the exon–intron boundary of exon 2 was used to exclude it during TFPI pre-mRNA processing. MDA-MB-231 cells were used for these experiments because preliminary studies demonstrated that they were amenable to Morpholino delivery, whereas EA.hy926 cells were not. Furthermore, MDA-MB-231 cells express TFPIα and TFPIβ at levels comparable with that in EA.hy926 cells.³⁹ To confirm successful exclusion of exon 2, the exon 1 to TFPIβ with nested exon 1 to exon 3 PCR was performed on cDNA made from Morpholino-treated cells. As a further step, PCR products were subjected to AvaII digestion, which cleaves within exon 2 to produce a 100-bp band. This band was observed in control Morpholino-treated cells (Figure 5B, lane 3) but not in exon 2 Morpholino-treated cells (Figure 5B, lane 6), confirming removal of exon 2 from TFPIβ. TFPIβ is the predominant TFPI isoform at the cell surface while TFPIα is secreted.²⁷ Therefore, cell surface TFPI, as a measure of TFPIβ expression, was assessed in Morpholino-treated cells by flow cytometry. Exon 2 Morpholino-treated cells had significantly greater cell surface TFPI compared with control Morpholino-treated cells (Figure 5C), confirming that exon 2 is a negative regulator of TFPIβ protein production. The observed increase is consistent with the percent TFPIβ mRNA that contains exon 2 (≈10%) in MDA-MB-231 cells. Furthermore, equal quantities of TFPIα were secreted into the media of control Morpholino-treated cells and exon 2 Morpholino-treated cells as measured by AlphaLISA (Figure 5D), confirming that exon 2 does not regulate TFPIα production.

Figure 5.

Suppression of exon 2 inclusion during tissue factor pathway inhibitor (TFPI) pre-mRNA splicing using a Morpholino increases TFPIβ but not TFPIα production in MDA-MB-231 cells. Ai, Schematic of the TFPI pre-mRNA comprising TFPIβ structural elements. The exon 2 Morpholino, designed to span the exon 2/intron 2 boundary, is depicted as the solid black line below exon 2. Dark gray arrows represent primers used to amplify TFPIβ in the first round of nested polymerase chain reaction (PCR), whereas light gray arrows represent primers used to amplify the region spanning exons 1 through 3 in the second round of nested PCR. Approximate positions of the AvaII restriction sites are indicated. ii, Schematic of possible TFPIβ mRNAs present in control Morpholino-treated cells. iii, Schematic of the probable TFPIβ mRNA present in exon 2 Morpholino-treated cells. B, MDA-MB-231 cells were treated with a Morpholino directed against the 3′ splice junction of TFPI exon 2 or a control Morpholino. The exon 1 to TFPIβ with nested exon 1 to exon 3 PCRs (Figure 1A and 1C) were performed, and the products were digested using AvaII. Lanes 1 and 4 are molecular weight markers, lanes 2 and 5 are undigested control reactions, and lanes 3 and 6 are digested reactions from cells treated with control or exon 2-specific Morpholinos. In lane 3, the lower band (100 bp) results from AvaII cleavage within exon 2, and the upper band (183 bp) results from cleavage between exons 6 and 7 within the coding region of TFPIβ. In lane 6, only the upper band is present because of the exclusion of exon 2, induced by the exon 2–specific Morpholino. The regions of TFPIβ mRNA correlating to each band are depicted in the right of the gel image. C, Cell surface TFPI expression on exon 2 or control Morpholino-treated MDA-MB-231 cells, as determined by flow cytometry. The mean fluorescence intensity (MFI) was calculated, the average MFI of the control Morpholino-treated cells designated as 100%, and individual MFIs normalized to this value. A statistically significant increase in cell surface TFPI expression on exon 2 Morpholino-treated cells was observed (P<0.05; n=4 for each Morpholino). D, TFPIα in the media of exon 2 or control Morpholino-treated MDA-MB-231 cells was determined by AlphaLISA and normalized to the total protein content in cell lysates. No difference in TFPIα production was observed between exon 2–treated cells or control-treated cells (n=4 for each Morpholino). For B and C, the bars represent the mean±1 SD. Mo indicates Morpholino.

Exon 2 Is a General Repressor of Protein Translation That Is Overcome by the TFPIα 3′ UTR

RNA is able to circularize, allowing interactions between regulatory elements within 5′ and 3′ UTRs.^40,41 Thus, exon 2 could repress TFPIβ translation by (1) acting synergistically with a specific element within the TFPIβ 3′ UTR, resulting in the repression of TFPIβ only, or (2) acting as a general repressor of translation that is overcome by the TFPIα 3′ UTR. To test these mechanisms, Ex1 or Ex1+Ex2 were paired with the β-actin 3′ UTR (Figure 6A), whose mRNA is not translationally regulated.³⁶ Compared with Ex1, Ex1+Ex2 had markedly decreased luciferase activity in CHO (79%; Figure 5B) and EA.hy926 (89%; Figure 6C) cells, demonstrating that exon 2 is a general negative translational regulatory element whose effects are overcome by the TFPIα 3′ UTR.

Figure 6.

Luciferase assays demonstrate that exon 2 is a general repressor of translation. A, Schematic of the luciferase constructs used. B and C, Luciferase assays were performed using Ex1-β actin and Ex1+X2-β actin chinese hamster ovary (CHO) (B) or EA.hy926 (C) cells. Luciferase activity in both cell lines was decreased in Ex1+X2-β actin cells compared with Ex1-β actin cells (P<0.01; n=6 for both cell lines). D and E, Luciferase assays were performed using Ex1-TFPIα Short and Ex1+X2-TFPIα Short CHO (D) or EA.hy926 (E) cells. Luciferase production was moderately decreased in Ex1+X2-TFPIα Short CHO cells compared with Ex1-TFPIα Short CHO cells (P<0.01), but no difference between the 2 constructs was observed in EA.hy926 cells (P>0.05; n=6 for both cell lines). For B through E, luciferase activity was determined as outlined in the Methods section, the average luciferase activity of all control construct replicates for that figure (the left-hand bar of each graph) designated as 100%, and individual luciferase activities compared with this value to give normalized luciferase activity as a percentage. Bars represent mean±1 SD. Ex1 indicates TFPI exon 1; Ex1+Ex2, TFPI exon 1 spliced to exon 2; TFPI, tissue factor pathway inhibitor; and UTR, untranslated region.

To determine whether the short TFPIα 3′ UTR generated via alternative polyadenylation is sufficient to overcome the repressive effect of exon 2, Ex1 or Ex1+Ex2 were paired with the 86-bp TFPIα 3′ UTR present in the 1.4-kb TFPIα message (Figure 6A). Compared with the Ex1-only construct, the Ex1+Ex2 construct displayed moderately reduced luciferase activity in CHO cells (46%; Figure 5D) but not in EA.hy926 cells (Figure 6E).

Discussion

Multiple TFPI transcripts arise from alternative splicing events that occur at the 5′ and 3′ ends of the human TFPI pre-mRNA.^8,20 Splicing within the 3′ region has been relatively well studied and produces the TFPIα and TFPIβ protein isoforms,⁸ which have important physiological differences in their structure and cellular expression, as well as the protease complexes they target to exert anticoagulant activity.^2,42 This study, using polysome analysis, luciferase reporter assays, and Morpholino exon-skipping experiments, describes for the first time that the 5′ splicing event involving exon 2 controls TFPIβ expression.

We initially hypothesized that the presence of exon 2 would tightly correlate with the 3′ splicing event that gives rise to either TFPIα or TFPIβ. However, Northern blot and nested PCR studies revealed that message for TFPIα and TFPIβ may contain or lack exon 2, demonstrating that the 5′ and 3′ splicing events occur independently. Studies by Pendurthi et al²⁹ suggested that changes in TFPI production by vascular smooth muscle cells in response to stimulation with serum or basic fibroblast growth factor occur via translational rather than transcriptional regulation. Therefore, polysome analysis and luciferase reporter assays were performed to examine whether exon 2 is involved in translational control of TFPI production. The results of these studies demonstrate that alternative splicing of exon 2 serves as a molecular switch that almost totally (≈90%) represses TFPIβ protein production. This degree of repression is somewhat surprising, given that a significant portion (≈50%) of TFPIβ message containing exon 2 is found in light polysomes. Because of the longer-than-average TFPI 5′ UTR, it may be that several 43S initiation complexes are loaded onto the TFPIβ message then stalled at exon 2, which may be sufficient to shift it into the light polysome fraction. Furthermore, Morpholino-based exclusion of exon 2 from TFPI mRNA in MDA-MB-231 cells increased TFPIβ on the cell surface but had no effect on TFPIα secreted from the cells, confirming that exon 2 is a negative regulator of endogenous TFPIβ but not TFPIα protein production.

Efficient translation is facilitated by mRNA circularization, which brings regulatory elements within 5′ and 3′ UTRs into close proximity and permits interactions between the 2.^40,41 In this context, it is interesting that exon 2 exerts its repressive effects when paired with the TFPIβ and β-actin 3′ UTRs, suggesting that exon 2 represents a general negative translational regulatory element. Potential repressive elements within 5′ UTRs include upstream open reading frames, upstream AUGs, or secondary structural elements that limit or prevent ribosome binding.^12,14 Theoretically, exon 2 contains 1 open reading frame at position 307 to 309 (Figure I in the online-only Data Supplement).²⁰ However, the only AUG initiation codon is immediately followed by a TGA stop codon and is unlikely to be repressive. Therefore, it is likely that secondary structural elements are responsible for the observed translation inhibition.

The TFPIα 3′ UTR contains 2 polyadenylation sites and, consequently, is produced in long (2700 bp) and short (86 bp) forms. Exon 2 did not substantially alter the production of luciferase constructs containing the TFPIα long 3′ UTR after transfection into either CHO or EA.hy926 cells, suggesting that the general repressive effects of exon 2 are overcome by the long TFPIα 3′ UTR. In experiments performed with luciferase constructs containing the short TFPIα 3′ UTR, exon 2 repressed luciferase production when transfected into CHO cells but not when transfected into EA.hy926 cells, suggesting that the shorter 86-bp TFPIα 3′ UTR contains at least a portion of the elements necessary for repression of the effects of exon 2. This finding is somewhat surprising, given that the long TFPIα 3′ UTR has the potential to contain many regulatory motifs. However, the presence of a derepressive element in this region confers resistance of exon 2–dependent translational inhibition to all forms of TFPIα message. This might be important for stable TFPIα protein expression in disease states where 3′ UTRs are shortened because of the selective use of proximal polyadenylation sites.¹⁷

A mechanism of translational control of alternatively spliced forms of estrogen receptor β involving 5′ and 3′ UTR interactions has recently been described,⁴³ in which different 5′ UTRs act with different isoform-specific 3′ UTRs to regulate basal and estrogen-mediated estrogen receptor β isoform expression. However, in contrast to the 2 exons comprising the TFPI 5′ UTR, the different estrogen receptor β 5′ UTRs are independent unique sequences found upstream of exon 1.⁴³ The current description of translational repression of a protein isoform (ie, TFPIβ) generated via alternative splicing at the 3′ end of a pre-mRNA by a second, independent splicing event (ie, inclusion of exon 2) at the 5′ end of the same pre-mRNA represents a previously unrecognized mechanism for translational control of protein synthesis.

Alternative splicing of exon 5 of the TF pre-mRNA determines whether TF is membrane anchored or soluble.⁴⁴ This splicing event is controlled by the serine/arginine-rich proteins ASF/SF2, SC35, SRp40, and SRp55.^45,46 Therefore, it is of interest to identify the factors involved in the TFPI exon 2 splicing event because they may control the balance of TF/TFPI expression. Preliminary in silico analysis of exon 2 identified putative exon splicing enhancer elements for the serine/ arginine-rich proteins ASF/SF2, SC35, and SRp40. Future biochemical studies are warranted to determine whether these factors are involved in the regulation of exon 2 splicing.

Translational regulation has been demonstrated to be important for spatial^47,48 and temporal⁴⁹ protein expression. There were minimal differences in the ratio of TFPIα mRNA lacking exon 2 to that containing it. In contrast, there were large variations in the amount of exon 2–containing TFPIβ mRNA present in different tissues. For instance, relative to TFPIβ mRNA lacking exon 2, high amounts of TFPIβ mRNA containing exon 2 (≤40%) were present in adult testis, thymus, brain, and lung, whereas relatively low amounts were observed in fetal liver and brain and adult heart, kidney, and liver. This suggests that one role of exon 2 might be tissue-specific regulation of TFPIβ expression. Although a physiological function for alternative splicing of exon 2 remains to be defined, it is proposed that different vascular beds maintain a latent pool of TFPIβ message that can be promptly translated in response to specific physiological stimuli that overcome the repressive effects of exon 2. Because TFPIβ has been demonstrated to be an effective inhibitor of both TF-dependent procoagulant and cellular migration events,⁹ it is tempting to speculate that factors involved in these processes might stimulate translation of this latent pool of TFPIβ message.

Methods

Northern Blot Analysis

³²P-labeled probes against TFPI exons 1, 2, 6 (total TFPI), 8 (TFPIβ), and 9 (TFPIα) were PCR amplified. Reactions (25 μL) contained 1× DyNAzyme II Hot Start Reaction Buffer, 200μM dATP, dGTP and dTTP, 2μM dCTP, 8.3nM ³²P-labelled dCTP (3000 Ci/mmole), 1μM forward and reverse primer (Table S I), 0.5U DyNAzyme II Hot Start Taq polymerase, and 0.25μL human placental cDNA. Cycling conditions were: 10m at 94°C; 30 cycles of 30s at 94°C, 30s at the relevant annealing temperature (Table S I), and 1m at 72°C; and 5m at 72°C. The probe was purified (QIAGEN PCR Purification kit, Valencia, CA) and specific activity determined before use.

Fifteen μg human lung RNA (Life Technologies, Grand Island, NY) was separated on a 1% formaldehyde gel (4mM MOPS, 1.2mM Na-Acetate, 2mM EDTA, 3% formaldehyde, pH 7.0) and transferred then UV-crosslinked to Biodyne B membrane. The membrane was pre-hybridized at 55°C in hybridization solution (0.34M Na₂HPO₄, 0.16M NaH₂PO₄, 7% SDS) before addition of ³²P-labelled probe (≈1 × 10⁶ cpm/mL), and incubated overnight at 55°C. It was washed twice at 23°C with 2× SSC (0.3M NaCl, 30mM Tri-Na-Citrate, pH 7.0), 0.1% SDS, then twice at 65°C with 0.2× SSC (30mM NaCl, 3mM Tri-Na-Citrate, pH 7.0), 0.1% SDS, before autoradiography at -80°C. Blots probed with the exon 1, 2, 6, or 9 probes were exposed to film for two weeks, while that probed with the exon 8 probe was exposed to film for three weeks.

Nested TFPIα and TFPIβ PCRs

Nested PCR was performed to examine exon 2 splicing in TFPIα and TFPIβ. The region spanning exon 1 to TFPIα or TFPIβ was first amplified using the Exon 1 Outside and TFPIα or TFPIβ primers (Table S II). Reactions (20μL) contained 1× Taq Pro Complete (2.0mM MgCl₂), 0.625μM forward and reverse primer, and 1μL human placental cDNA. Cycling conditions were: 3m at 94°C; 35 cycles of 30s at 94°C, 30s at 57°C, and 1m 30s at 72°C; and 5m at 72°C. In the nested reaction, Exons 1 through 3 were amplified using the Exon 1 Inside and Exon 3 primers (Table S II). Reactions (20μL) contained 1× Taq Pro Complete (2.0mM MgCl₂), 0.625μM forward and reverse primer, and 1μL of product from the first reaction. Cycling conditions were: 3m at 94°C; 5 cycles of 30s at 94°C, 30s at 67°C decreasing to 62°C, and 45s at 72°C; 15 cycles of 30s at 94°C, 30s at 62°C, and 45s at 72°C; and 5m at 72°C. Products were separated on a 4% agarose gel, the bands isolated by gel extraction (QIAquick Gel Extraction kit, Valencia, CA), and the sequence determined. In Morpholino experiments, some nested PCR products were digested with AvaII prior to separation on a 4% agarose gel.

Tissue cDNA Analysis

cDNA was produced from 1μg RNA (Human total RNA, Master Panel II, BD Biosciences, San Jose, CA) using Superscript II Reverse Transcriptase. The exon 1 to TFPIβ PCR, as outlined as the initial PCR in the nested TFPIα and TFPIβ PCRs section, was performed and the reactions separated on a 1.5% agarose gel. Gel images were obtained and analyzed using AlphaImager HP, version 3.4.0, ensuring that images weren't saturated before analysis, the background corrected band intensity determined, and the ratio of TFPIα or TFPIβ mRNA lacking exon 2 to the corresponding mRNA containing exon 2 calculated.

Polysome Isolation and Analysis

2 × 10⁷ HUVECs were lysed in 1mL lysis buffer (0.2M Tris, pH 9.0, 0.2M KCl, 25mM EGTA, 50mM MgCl₂, 1% NP-40, 0.5% Na-deoxycholate, 500U/mL RNasin, 5mM Dithiothreitol, 50μg/mL Cycloheximide, 50μg/mL Chloramphenicol, 0.5mg/mL heparin, 1mM AEBSF) and the lysate clarified by centrifugation at 12 000 × g for 5m at 4°C. The supernatant was loaded onto a 10mL 20-60% sucrose gradient (sucrose in 50mM Tris-HCl, pH 8.4, 25mM KCl, 5mM MgCl₂, 5mM Dithiothreitol, 50μg/mL Cycloheximide, 50μg/mL Chloramphenicol, 0.5mg/mL heparin) and separated by ultracentifugation at 247 000 × g for 1.5h at 4°C in an SW41-Ti rotor. Fractions (500μL) were collected into 10μL 0.5M EDTA and 20μL RNAsecure (Life Technologies, Grand Island, NY), heated for 10m at 60°C, and A₂₆₀ determined. Proteinase K (200μg/mL, New England Biolabs, Ipswich, MA) was added and incubated at 37°C for 30m. RNA was isolated by phenol/chloroform extraction, precipitated using lithium chloride (2.5M LiCl, 17mM EDTA), pelleted by centrifugation at 20 000 × g for 15m at 4°C, washed with 70% ethanol, and resuspended in 20μL TE Buffer (10mM Tris-HCl, 1mM EDTA, pH 7.5). cDNA was produced from 100ng RNA using Superscript II Reverse Transcriptase (Life Technologies, Grand Island, NY), according to manufacturers' instructions. The distribution of β-actin, TGF-β1, TFPIα, and TFPIβ message was determined using the initial reaction of the Nested TFPIα and TFPIβ PCR. Primers for β-actin and TGF-β1 are listed in Table S II. All reactions were separated on a 1.5% agarose gel and analysed by densitometry, using ImageJ (NIH, Bethesda, MD, version 1.46r). The amount of specific product (i.e. exon 2 containing TFPIβ) in each fraction was expressed as a percentage of the amount of that same product in the entire gradient.

Production of Luciferase Reporter Constructs

The pCMV-GLuc vector (New England Biolabs, Ipswich, MA) was modified to remove the poly(A) signal and to express GFP. To remove the poly(A) signal, the vector was digested with NotI and DraIII and an AscI site introduced by ligation using the AscI complementary oligonucleotides (Table S III, comp. oligo I and II), producing the pCMV-GLuc-AscI vector. For GFP expression, the IRES from pWPI and the copepod GFP (copGFP) from pCDH-CMV-MCS-EF1 -copGFP were PCR amplified and ligated into the pCMV-GLuc-AscI vector at the BstI and PfoI sites. This vector was named pCGAIC. Both exon 1 (Ex1) and exon 1 spliced to exon 2 (Ex1+Ex2) were amplified from human placental cDNA and ligated into the pCGAIC vector at the BamHI site. The 3′ UTR inserts were amplified from human genomic DNA and ligated into the pCGAIC vector at the NotI/AscI site. All inserts were amplified by PCR (Phusion High Fidelity PCR kit, New England Biolabs, Ipswich, MA), using primers listed in Table S III, cloned into the pJet 1.2 vector (CloneJet PCR Cloning Kit; Thermo Fisher, Waltham, MA), and the sequence verified before addition to pCGAIC.

Production of Stable Cell Lines Expressing Luciferase

CHO-K1 cells were cultured in F12K medium containing 10% fetal calf serum (FCS), 100U/mL penicillin, and 100μg/mL streptomycin. 4 × 10⁵ CHO cells were transfected using the Trans-IT CHO transfection kit (Mirus Biosciences, Madison, WI), stable transfectants selected using 800μg/mL G418, and GFP positive cells collected by fluorescence-activated cell sorting.

EA.hy926 cells were cultured in DMEM containing 10% FCS, 100U/mL penicillin, and 100μg/mL streptomycin. 3 × 10⁵ EA.hy926 cells were transfected using Lipofectamine LTX (Life Technologies, Grand Island, NY) and stable transfectants selected using 1mg/mL G418.

Luciferase Assays

3 × 10⁵ CHO cells were seeded in 35mm dishes in 2mLs of culture media and incubated for 48h, after which the media was collected, centrifuged at 20 000 × g for 15m at 4°C, and the supernatant frozen at -20°C until assay. Luciferase activity was measured in 25μL of sample using 50μL of luciferase substrate (BioLux Gaussia Luciferase assay kit, New England Biolabs, Waltham, MA), with a 4s delay time and 5s counting time. Cells were washed with PBS (4.8mM Na₂HPO₄, 0.7mM KH₂PO₄, 137mM NaCl, 2.7mM KCl, pH 7.4), harvested by scraping in PBS with 10mM EDTA, separated into 400μL and 100μL aliquots, centrifuged at 500 × g for 5m at 4°C, and washed. The 400μL aliquot was resuspended in 250μL of PBS, duplicate 100μL volumes dispensed into the wells of a black 96-well plate, and centrifuged for 5m at 500 × g. GFP production was assessed by fluorometry, using a 10s read at 485/535 nm. The 100μL aliquot was lysed in 62.5μL lysis buffer (PBS with 30mM CHAPS, 10mM EDTA, 50μM 3,4-DCI, 100μM E-64), centrifuged at 20 000 × g for 15m, and the lysate protein concentration determined (BCA Protein Assay, Pierce, Rockford, IL).

Luciferase activity in EA.hy926 cells was determined as for CHO cells, except that 1 × 10⁵ EA.hy926 cells were initially seeded, and luciferase activity was normalized to luciferase mRNA. RNA was isolated (QIAGEN RNeasy Mini kit, Valencia, CA), pooled from triplicate cultures, 5μg Turbo DNase treated, and 400ng used for cDNA production using Superscript II Reverse Transcriptase. Real-time PCR was performed on the Applied Biosystems 7500 PCR system, using Power SYBR Green and the primers listed in Table S IV.

To allow for comparison between CHO and EA.hy926 cells, for each 3′ UTR the luciferase activity for each replicate was normalized to the average luciferase activity of the control construct, which in most cases was Ex1 plus the 3′ UTR.

Morpholino Treatment of MDA-MB-231 Cells

3 × 10⁵ MDA-MB-231 cells were seeded in 35mm dishes and grown to 80-100% confluence in DMEM containing 10% FCS, 100U/mL penicillin, and 100μg/mL streptomycin. Morpholinos (10 μM; TFPI Exon 2 sequence: 5′-ACAGAAATTTGTATCTCACAGTTCT-3′; Control sequence: 5′-CCTCTTACCTCAGTTACAATTTATA-3′) were delivered into cells using 6 μL of Endo-porter delivery reagent (GeneTools, Philomath, OR), in DMEM containing 10% FCS. After 24 hours, media was collected, centrifuged at 500 × g for 10 minutes at 4°C, the supernatant collected and centrifuged at 20 000 × g for 10 minutes at 4°C, and the supernatant collected then frozen at -20°C. Cells were washed with PBS, harvested by scraping in PBS with 10mM EDTA, centrifuged at 500 × g for 5m at 4°C, and washed. Cells were then processed for RNA isolation and cDNA production (as outlined in ‘Luciferase Assays’) or flow cytometry.

AlphaLISA

An in-house AlphaLISA was established to quantitate TFPIα in culture media. A mouse anti-human TFPI antibody, directed toward the second Kunitz domain (anti-K2; kindly donated by Novo Nordisk, DK), was conjugated to AlphaLISA acceptor beads (PerkinElmer, Waltham, MA), according to manufacturers' instructions. Sample (5 μL) was incubated for 1 hr at 37°C with a 20 μL mixture of 12.5 μg/mL anti-K2 beads and 5 nM biotinylated anti-human TFPI antibody, directed toward the third Kunitz domain (kindly donated by Novo Nordisk, DK). Streptavidin-conjugated AlphaLISA acceptor beads (25 μL, 40 μg/mL; (PerkinElmer, Waltham, MA) were then added and incubated for 1 hr at room temperature. Plates were read using the default AlphaLISA protocol on an Envision plate reader (PerkinElmer, Waltham, MA) and TFPI concentration determined using a standard curve generated with human recombinant TFPI (kindly donated by Novo Nordisk, DK).

Flow Cytometry

All incubations were performed on ice. Cells were blocked for 30m with a 1:1 mixture of protein-free blocking reagent and 1% casein (P:C block; both from G-Biosciences, St Louis, MO) containing 10% Fc receptor blocker (Innovex Biosciences, Richmond, CA) and 2.5% normal goat serum (NGS; Jackson Immunoresearch, West Grove, PA). They were then incubated with rabbit anti-human TFPI antibody (Sekisui Diagnostics, Stamford, CT; 1:100 in P:C block with 2.5% NGS) for 1 h, washed 3× with PBS, incubated for 30m with Alexa Fluor® 488-conjugated (Fab')₂ goat anti-rabbit IgG (1:500 in P:C block; Invitrogen), then washed 3× and resuspended in P:C block. Cells were analysed using an Accuri C6 Flow Cytometer (BD Biosciences, Franklin Lakes, NJ) and FlowJo software, version 7.6.5 (Tree Star, Ashland, OR)

Statistical Analysis

The students' t -test was used to test for statistical significance, using GraphPad Prism for Windows, v4.0 (Graphpad Software, San Diego, CA). A p-value less than 0.05 was considered statistically significant.

Supplementary Material

Figure S I. Sequence of Exons 1 through 3 of TFPI mRNA. Exons 1 and 3 are depicted in regular font, while exon 2 is highlighted grey and italicized. Exon 1 begins at nucleotide (nt) 1, exon 2 at nt 294, and exon 3 at nt 416. Potential translation start codons are in larger font, bolded, and underlined. Exon 2 does contain an ATG translation start codon (nucleotide position 307-309) that is immediately followed by a TGA stop codon. Exon 3 contains two potential translation start sites (nucleotide positions 418-420 and 430-432). Neither contain an optimal Kozak consensus sequence [gcc(a/g)ccAUGG]. However, the 430-432 site does have an ‘a’ at the -3 position (numbering relative the translation start site) and an ‘a’ at the +1 position and is therefore likely the most commonly used translation start site.

Significance.

Two tissue factor pathway inhibitor (TFPI) isoforms, TFPIα and TFPIβ, are produced via alternative splicing toward the 3′ end of the TFPI gene. A separate splicing event within the 5′ untranslated region produces TFPI mRNA lacking exon 2. We demonstrate that exon 2 is a negative translational regulatory element that represses production of TFPIβ but not TFPIα. This is the first description of a 5′ untranslated region splicing event regulating translation of protein isoforms produced via a second, independent splicing event. Variable expression of TFPIβ mRNA containing exon 2 across a range of human tissues suggests that one role of exon 2 splicing in TFPIβ message is tissue-specific control of TFPIβ protein expression.

Nonstandard Abbreviations and Acronyms

Ex1

TFPI exon 1

Ex1+Ex2

TFPI exon 1 spliced to exon 2

PCR

polymerase chain reaction

TF

tissue factor

TFPI

tissue factor pathway inhibitor

UTR

untranslated region