The selenium transport protein, Selenoprotein P, requires coding sequence determinants to promote efficient Selenocysteine incorporation

Abstract

Selenoproteins are an essential and unique group of proteins in which selenocysteine (Sec) is incorporated in response to a stop codon (UGA). Reprograming of UGA for Sec insertion in eukaryotes requires a cis-acting stem loop structure in the 3’ untranslated region of selenoprotein mRNA and several trans-acting factors. Together these factors are sufficient for Sec incorporation in vitro, but the process is highly inefficient. An additional challenge is the synthesis of Selenoprotein P (SELENOP), which uniquely contains multiple UGA codons. Full SELENOP expression requires processive Sec incorporation, the mechanism for which is not understood. In this study, we identify core coding region sequence determinants within the SELENOP mRNA that govern SELENOP synthesis. Using ⁷⁵Se labeling in cells, we determined that the N-terminal sequence (upstream of the 2nd UGA) and C-terminal sequence context are two independent determinants for efficient synthesis of full length SELENOP. In addition, the distance between the first UGA and the consensus signal peptide is also critical for efficiency.

INTRODUCTION

Selenoproteins constitute a small but important group of 25 polypeptides that are primarily known for their role in conferring protection against oxidative stress in humans, reviewed in (1, 2). Although diverse in their structure and function, selenoproteins share at least one Sec residue in their amino acid sequence, which is encoded by an in-frame UGA codon. Thus, UGA, which is typically a translation termination signal, is recoded for Sec incorporation in selenoproteins. In addition to the classical translation factors, selenoprotein synthesis requires both cis-acting sequences and trans factors. The cis-acting sequences include an in-frame UGA codon and a stem-loop in the 3’ untranslated region (UTR) of all eukaryotic selenoprotein mRNAs referred to as selenocysteine insertion sequence (SECIS) and the trans factors are3 the SECIS binding protein (SBP2), a specialized elongation factor (eEFSec) and the selenocysteine tRNA (Sec-tRNA^Sec) (3–8). While these factors are sufficient for Sec incorporation in vitro, it occurs very inefficiently where ~90% of the time the UGA codon is interpreted as a termination codon (9–11). The abundant expression of several selenoproteins in animals, however, indicates that Sec incorporation is likely to be efficient in vivo, suggesting the existence of an efficiency factor or element (12–14).

Several regulatory factors have been found to differentially affect selenoprotein expression both in vitro and in cells. These include novel SECIS binding proteins, Se availability, UGA codon position, UGA codon context, distance between the UGA and the SECIS element, mRNA stability and cis-elements located both in the coding region and 3’UTR (4), (15–24). Almost all of these factors have been studied in the context of selenoproteins with a single Sec codon.

A remarkable exception to the single Sec selenoprotein is selenoprotein P (SELENOP). It is the only selenoprotein that incorporates multiple Sec residues in its primary sequence (25, 26), and its mRNA possesses two SECIS elements in the 3’ UTR (4, 27). Depending on the species, SELENOP contains 10–35 Sec residues requiring processive recoding of UGA codons. Selenium-rich SELENOP is synthesized mainly in hepatic cells and secreted into the blood (28,29) where it serves as a transport protein for selenium to extrahepatic tissues (30). Thus, full-length SELENOP synthesis requires processive Sec incorporation from a single mRNA. Very little is known to date about processive Sec incorporation.

Processive incorporation of multiple Sec residues is known to be a mechanistic bottleneck because it cannot be efficiently replicated in artificial reporter systems in vitro (19, 31). However, both cellular expression and in vitro translation of native SELENOP results in processively made full length product, although termination at the first and second UGA codon is commonly observed (32–34). An in vitro study using dual reporter constructs analyzed Sec incorporation efficiency of 2–3 UGA codons placed in native context versus 2–4 UGA codons placed in a random sequence context. In both constructs Sec incorporation at the first UGA was 10-fold less efficient than at downstream UGA codons and this result was unaffected by UGA sequence context (34). This raises a fundamental mechanistic question about processive Sec incorporation: What differentiates Sec incorporation at the first UGA versus downstream UGA codons? Recently a direct comparison of in vitro synthesis of wild type human SELENOP versus an artificial luciferase construct with 10 UGA codons positioned exactly as found in human SELENOP but placed in a non native context was performed (35). In this case, the artificial selenoprotein resulted in dramatically lower efficiency and processivity, thus strongly implicating a role for coding region sequence. Furthermore, functional polymorphisms located in the coding region of the human SELENOP gene have been found to regulate the distribution of full length versus early termination products (36). Also, bioinformatic analysis identified four structural elements in the coding region termed as initiation stem loop (ISL) and Sec Redefinition Elements -SRE1, SRE2 and SRE3. In a reporter construct, the ISL and SRE1 influenced translation initiation and Sec incorporation at the first UGA, respectively while the role of SRE2 and 3 are yet to be determined (35). And finally, since all artificial constructs with multiple Sec codons have been analyzed using SELENOP SECIS elements, it seems most likely that although SECIS elements are sufficient for Sec incorporation, the coding region sequence is critical to modulate efficiency and processivity. However, to date, there exists no systematic functional analysis on the coding region sequence determinants in their native contexts.

Thus, the primary aim of this work was to identify and examine determinants within SELENOP mRNA that regulate processive and efficient SELENOP synthesis. We examined the coding region for regulatory elements both in a cellular environment and in vitro by creating in-frame deletion mutants. We also individually assessed the roles of the N- and C-terminal portions of SELENOP for their roles in multiple Sec incorporation. Interestingly, we have found that in cells the SELENOP N-terminus can remotely drive processive Sec incorporation of downstream Sec codons that are placed in a non-native sequence environment. We also demonstrate that C-terminal sequence context can independently drive processivity from 2nd - 10th UGA but only in the absence of the first UGA. On the other hand, the distance between the first UGA and the signal peptide is a determinant of efficiency. In parallel, we have also identified that the zebrafish SELENOP can be used as a robust system to identify determinants of processive Sec incorporation and that expression of mammalian SELENOP is governed by additional determinants.

RESULTS

Expression of SELENOP in mammalian cells.

To identify sequence determinants that regulate processive Sec incorporation in a cellular environment, we set out to establish a SELENOP expression system in cultured cells. To do this, we transfected either human (10 Sec residues) or zebrafish (17 Sec residues) SELENOP cDNA into HEK293 cells (HSELENOP or ZSELENOP). For comparison, we also created cysteine versions of SELENOP constructs where all Sec residues were changed to Cys (HCysSELENOP and ZCysSELENOP). A FLAG tag was added to the C-terminus of all transfected constructs (diagrammed in Fig 1A). Stably transfected cell lines were monitored for SELENOP expression by incubating cells with ⁷⁵Se-selenite followed by analysis of the conditioned media by SDS-PAGE (Fig 1B). Full length HSELENOP has a theoretical molecular weight of 44 kDa, but when derived from conditioned medium it migrates at an apparent molecular weight of ~63kDa due to glycosylation (36, 37). Thus as expected, ⁷⁵Se labeling of un-transfected HEK293 cells (Fig 1B, lane 1) produced two endogenous HSELENOP bands that migrate between 50 and 70 kDa. Earlier studies have confirmed that the two HSELENOP isoforms result from a difference in length of polypeptide chain and not due to variation in glycosylation pattern (36, 38, 39)(38–40). The top band corresponds to full length protein while the lower band is an early termination product corresponding to termination at the second Sec codon. After correcting for the number of Sec residues, we observe a 1:5 ratio of the full length to early termination product for endogenous HSELENOP indicating preferential production of the truncated isoform in these cells. HEK293 cells stably transfected with ZSELENOP produced an intense band at 46 kDa corresponding to full length protein and a weak band at 38 kDa (40). The apparent size difference between human and zebrafish SELENOP is likely due to the fewer prediced glycosylation sites in ZSELENOP (40). Similar to HSELENOP, transfected ZSELENOP displayed a 1:3.7 ratio of the full length to early termination product in these cells. In addition, ZSELENOP-expressing cells show small but significant reduction of endogenous full length HSELENOP (Fig 1B, lane 2 and Fig 1C).

FIGURE 1. Stably transfected ZSELENOP can be expressed in HEK293 cells.

A) Schematic representation of SELENOP cDNA transfected into HEK293 cells. Black vertical lines within the coding region of SELENOP mRNAs represent the Sec codons. A FLAG tag sequence was inserted into the 3’ end of each construct. B) Untransfected and G418 selected stable cell lines transfected with either ZSELENOP, ZCysSELENOP, HSELENOP or HCysSELENOP analyzed for SELENOP expression using ⁷⁵Se labeling. Arrows indicate ⁷⁵Se labeled full length and early termination product of endogenous HSELENOP and transfected ZSELENOP. Media from stable cell lines were resolved by SDS-PAGE and detected by PhosphorImager analysis. C) Graphical representation of quantitated variations in endogenous HSELENOP isoform expression data from B. D) anti-Flag Western analysis of conditioned media from cells stably expressing ZSELENOP, ZCysSELENOP, HSELENOP or HCysSELENOP E) qRT-PCR analysis of mRNA data obtained for samples from panel B). Equal amounts of total RNA was used for qRT-PCR. Each sample was evaluated in triplicate. The fold increase in either HSELENOP or ZSELENOP mRNA relative to actin was determined by the comparative CT method. Data is plotted as the average plus and minus standard deviation for three biological replicate experiments. P < 0.05 compared to value with the untransfected using paired two-tail students t-test.

Surprisingly, stable transfection of human SELENOP cDNA did not result in any detectable increase in total human SELENOP production at 63 kDa (Fig 1B, lane 4) compared to untransfected cells. Instead, we observed a ~ 2-fold increase in the early termination product of SELENOP with no significant change in full length SELENOP production (Fig 1B, lane 4 and Fig 1C). A similar effect on SELENOP expression was observed with transfection of HCysSELENOP cDNA (Fig 1B lane 5 and Fig 1C). These results suggest that overexpression of transfected human SELENOP or CysSELENOP, may titrate mammalian-specific factors thereby increasing early termination. Since CysSELENOP variants cannot be detected by ⁷⁵Se labeling, we immunoblotted the gel with anti-FLAG antibody. We confirmed expression of CysSELENOP for both human and zebrafish, but HCysSELENOP expression was ~3-fold lower than ZCysSELENOP (Fig 1D). We were unable to detect any of the wild type (Sec-containing) ZSELENOP by FLAG western on media, indicating lower translation efficiency compared to its cysteine version.

To determine if the apparent inability to express transfected human SELENOP was due to differences in mRNA levels, we performed qRT-PCR. As shown in Fig 1E, the level of ZSELENOP mRNA in stably transfected cells was 9-fold higher than endogenous HSELENOP mRNA levels. In contrast, HSELENOP cDNA transfection resulted in only 3.5-fold higher mRNA levels than in untransfected cells. Similarly, ZCysSELENOP and HCysSELENOP mRNA was 25-fold and 9-fold higher than endogenous HSELENOP mRNA levels, respectively. Consistent with these results, rat SELENOP also cannot be expressed in other cell lines (data not shown). The mechanistic basis for the lack of mammalian SELENOP expression in mammalian cells remains unknown.

Full length expression of ZSELENOP in transfected mammalian cells-

In order to validate the use of ZSELENOP to study the mechanism of processive Sec incorporation, we set out to verify that the 46 kDa band detected in ⁷⁵Se-labeling corresponded to full length ZSELENOP. Since we were unable to detect wild-type ZSELENOP by anti-FLAG immunoblot (Fig 1D), we examined the extent to which the 46 kDa band for ZSELENOP expressed in cells was full length by immunoprecipitation. The C-terminal selenocysteine residues of ZSELENOP are closely spaced and detection of early termination at these residues is not possible under the experimental conditions used. Therefore we performed FLAG purification with anti-FLAG antibody from concentrated conditioned media using the C-terminal FLAG tag of transfected ZSELENOP to determine the fraction of ⁷⁵Se-labeled ZSELENOP that is full length. For comparison, we also performed FLAG purification on media from cell lines stably expressing non-FLAG-tagged ZSELENOP. Fig 2A lane 2, shows that immunoprecipitated full length ZSELENOP is detectable both by western blot and phosphorimage. To evaluate what percentage of the product is full length, we analyzed the supernatant of the FLAG immunoprecipitation (IP). As shown in Fig 2B, lanes 2 and 4, most of the product (80%) observed is full length ZSELENOP. Furthermore, to analyze if the band at ~38 kDa for ZSELENOP is a result of early termination or difference in glycosylation pattern, we deglycosylated ZSELENOP with PNGase F. Deglycosylation of ZSELENOP decreases the apparent molecular weight of both the 46 kDa and 38 kDa band by 1.5 and 2.6 kDa respectively (Supplementary Fig 1A). This result is in agreement with an earlier study which also showed that the mass difference between the two ZSELENOP translation products is due to distinction in the length of their polypeptide chain (40). We also identified that the early termination product for ZSELENOP corresponds to termination at the 2nd UGA by using a mutant (ZSELENOP_U59C) where the first UGA was mutated to UGC (Supplementary Fig1B). Overall, the data presented in Fig 1 and 2 and previous studies (32) establish that the expression of ZSELENOP in HEK-293 cells is a robust system to study the mechanism of processive SELENOP synthesis and can be used as a surrogate for the study of SELENOP translation in transfected cells.

FIGURE 2. Full length ZSELENOP can be expressed in HEK293 cells.

A) Top- anti-FLAG Western blot analysis of FLAG-IP eluant purified from concentrated media collected from ⁷⁵Se labeled stable cell lines expressing ZSELENOP with or without a FLAG tag. Bottom-, PhosphorImage of western blot from top panel B) PhosphorImage of FLAG ZSELENOP immunoprecipitated from stable cell line expressing wild type ZSELENOP.

SELENOP N-terminus coding region is sufficient for efficient Sec incorporation in cells-

In vitro translation of an artificial selenoprotein that contained multiple Sec codons in a non-native context and chimeric SELENOP mRNAs indicated that efficient Sec incorporation requires both N- and C- terminus sequence for optimal expression thereby suggesting a role for UGA codon context in efficiency (35). To examine if this is also true in a cellular environment, we tested the artificial selenoprotein and created chimeras similar to those used in the previous study. As shown in Fig 3A, the artificial selenoprotein Luc10UGA is a truncated version of firefly luciferase containing 10 Sec codons spaced exactly as found in human SELENOP. Luc10UGA also contains the signal peptide, the full SELENOP 3′ UTR, and a C-terminal FLAG tag. The N-terminal chimera (ZN-term_Luc) was made by fusing the N-terminal sequence of wild type ZSELENOP with the C-terminal sequence of Luc10UGA (Fig 3A 35). Similarly, we also fused the ZSELENOP C-terminal sequence to the artificial N-terminus sequence of Luc10UGA (Luc_ZC-term, Fig 3A). HEK293 cells stably or transiently transfected with these constructs were labeled with ⁷⁵Se followed by analysis of the conditioned media and cell lysates by SDS-PAGE. Differences in RNA abundance was analyzed using qRTPCR. As shown in Fig 3B lane 2, cells transiently expressing the artificial construct Luc10UGA yielded no expression in the media although a very weak band at ~50 kDa can be seen in the lysate. Meanwhile, the ZN-term_Luc chimera produced a distinct band at ~46 kDa in the media and in the lysate. Upon normalization for transfection efficiency and RNA levels, the overall efficiency of the ZN-term_Luc chimera for full length product was about two fold lower than wild-type ZSELENOP (Supplementary Fig 2A). In contrast, no full length product was detectable for Luc_ZC-term in the media but in the lysate a weak trail of bands was detectable. qRT-PCR analysis of these chimeric mRNAs showed that both ZN-term_Luc and Luc_ZC-term RNA levels were higher than wild type, while Luc10UGA was slightly reduced by ~2 –fold (Fig 3C). Stable expression of the chimeras produced similar results (data not shown).

FIGURE 3. Native N- terminal SELENOP sequence alone is sufficient for processive Sec incorporation downstream in cells while C-terminal sequence is not.

A) Schematic diagram of the SELENOP N- terminal and C-terminal coding region chimeras of ZSELENOP. B) ⁷⁵Se labeled media and lysate respectively from transient cell lines expressing coding region chimeras, resolved by SDS-PAGE and detected by PhosphorImager analysis. The arrows indicate full-length and early termination products. C) qRT-PCR analysis of mRNA data obtained for samples from B. Equal amounts of total RNA were used for qRT-PCR and the fold increase in ZSELENOP chimeras mRNA relative to wild-type ZSELENOP was determined by the comparative CT method. Data is plotted as the average plus and minus standard deviation for three independent experiments. P < 0.05 compared to value with the native ZSELENOP using paired two-tail students t-test. D) FLAG western blot of media and lysate from HEK293 cells expressing cysteine version of the constructs in A. E) qRT-PCR analysis of mRNA data obtained for samples from D as described in Fig 1B.

To test if the differences in the expression of N- and C- terminal chimeras are due to the artificial nature of the protein, we tested cysteine versions of all constructs shown in Fig 3A with FLAG tag on the C-terminus. As shown in Fig 3D, ZN-term_Luc_Cys chimera was detectable both in the media and lysate similar to its Sec-containing variant. The Luc10UGC and Luc_ZC-term_Cys were abundantly detectable in the lysate and migrated relatively slower compared to wild-type ZSELENOP that migrates at 46 kDa.

The fact that the artificial selenoprotein Luc10UGA was expressed inefficiently confirms prior studies that the native SELENOP coding region is crucial for efficiency of full length product in vitro (35). In addition, in transfected cells the N-terminal chimera (ZN-term_Luc) is sufficient to stimulate Sec incorporation efficiency at downstream UGA even in the absence of the native C- terminus SELENOP sequence, albeit with ~2-fold lower efficiency. However, in contrast to in vitro analysis (35) the C-terminus alone variant (Luc_ZC-term) is defective for both efficient and processive Sec incorporation.

SELENOP coding region sequence (58–165 nts) influences Sec incorporation efficiency-

In order to identify specific determinants required for efficient production of full-length SELENOP, we created consecutive in-frame deletions within the ZSELENOP coding region as shown in Fig 4A. All coding region deletions retained the signal peptide in the N-terminus and the FLAG tag on the C-terminus. Among the 7 coding region mutants, three of the deletions resulted in loss of Sec codons. This includes the N-terminal deletion Δ166–315, which resulted in loss of the first UGA and the C-terminal deletions Δ766–915 and Δ916–1101, which eliminated 6 and 10 Sec codons, respectively (Fig 4Atop). Stable and transient cell lines expressing cDNA of coding region deletion mutants were created. Both media and lysate as described above were analyzed post ⁷⁵Se labeling.

FIGURE 4. In-frame deletions within SELENOP coding region in transiently expressing cell lines.

A) Top panel, Schematic diagram of the SELENOP coding region deletion mutants. Bottom panel, ⁷⁵Se labeled media from transiently transfected cell lines expressing coding region deletion mutants resolved by SDS-PAGE and detected by PhosphorImager analysis. The arrow indicates full-length and early termination products for both native ZSELENOP and coding region deletion mutants. B) ⁷⁵Se labeled cell lysate from transients expressing coding region deletion mutants resolved by SDS-PAGE and detected by PhosphorImager analysis. The arrow indicates full-length and early termination products for both native ZSELENOP and coding region deletion mutants. C) PhosphorImage of FLAG ZSELENOP immunoprecipitated cell lines expressing wild type ZSELENOP and deletion mutants Δ766–915 and Δ916–1101.D) qRT-PCR analysis of mRNA data obtained for samples from panel A) as described in Fig 1B. E) Efficiency expressed as a ratio of protein band intensity in media and lysate to fold change in RNA for the ZSELENOP N-terminal chimera relative to wild type. F) Flag Western Blot analysis of media and lysate of cells transiently transfected with cysteine versions of wild type zebrafish and N-terminal deletion mutant Δ58–165. Both these constructs were flag tagged on the C-terminus. G) qRT-PCR analysis of mRNA data obtained for samples from panel E as described in Fig 1B.

As shown in Figure 4A, all N-terminal deletions (Δ166–315, Δ616–765, Δ316–465 and Δ466–615) showed varying levels of expression with the exception of Δ58–165 mutant which resulted in no detectable product in the media. To examine the effect of the deletions on full length efficiency, we quantitated the band intensity of full length product in both lysate and media and normalized it to transfection efficiency and RNA abundance (Fig 4B). As shown in Fig 4C, two mutants show significant reduction in efficiency. Δ58–165 expression levels were reduced almost 5-fold relative to wild-type ZSELENOP while Δ766–915 showed a 2-fold reduction. To distinguish if the reduced expression of Δ58–165 mutant was specific to Sec incorporation or affected total translation, we tested the cysteine version of Δ58–165 mutant. As shown in Fig 4D and E, both expression and RNA levels for cells transiently expressing ΔCys58–165 mutant were comparable to the wild type ZCysSELENOP. These results suggest a role for the 58–165 nt sequence in regulating of multiple Sec incorporation efficiency.

The native ZSELENOP coding sequence contains a consensus N-glycosylation site at amino acid position 109, thereby resulting in the observed molecular weight of 46 kDa. As expected, one of the deletion mutants, Δ316–465, migrated faster than others since it lacks the glycosylation site and thus was not secreted into the media (Fig 4A, lane 4). Interestingly, while the Δ58–165 and Δ166–315 mutants migrated at the expected molecular weight based on the deletion size, Δ466–615 migrated similar to the glycosylation deficient mutant Δ316–465, suggesting additional glycosylation sites in the 466–615 sequence (Fig 4A compare lanes 2–5). Among the C-terminal deletions, Δ766–915 migrated similarly to the native ZSELENOP while Δ916–1101 migrated faster (Fig 4A compare lane 7 and 8). In order to rule out the possibility that the faster migration of the Δ916–1101 product is the result of early termination we performed FLAG immunoprecipitation (IP). As shown in Fig 4F, wild-type, Δ766–915 and Δ916–1101 constructs all produced ~80% full length protein. It is also interesting to note from this data that the Δ166–315, Δ316–465 and Δ466–615 displayed a trail of early termination products in the lysate (Fig 4A), suggesting that these sequences may be influencing processivity at downstream UGAs or protein stability. All together these data reveal specific functionality for discrete regions of the ZSELENOP coding region, especially in the 58–185 region.

The distance between the first UGA and the signal peptide is important for efficiency.

The 58–165 nts sequence lies between the first UGA and the consensus signal peptide and is highly conserved. The deletion mutant Δ58–165 alters the distance between the first UGA and the start codon/signal peptide. To examine if the effect of the Δ58–165 mutant is due to the highly conserved nature of the sequence or because it affects the distance of the first UGA from the start codon/signal peptide, we created two mutants. As shown in Fig 5A the wild-type ZSELENOP sequence at 58–165 was replaced with sequence from Luc10UGA to create Z_58–165Luc. Similarly, in construct Luc_58–165Z the nucleotide sequence 58–165 in Luc10UGA was replaced with sequence from ZSELENOP. If the 58–165 nt sequence contains efficiency elements, then Luc_58–165Z is expected to show increased efficiency but if the distance is important then Z_58–165Luc will express efficiently relative to Δ58–165 mutant. These constructs were transiently transfected and the media and lysate analyzed for expression by ⁷⁵Se labeling (Fig 5B) and qRT-PCR (Fig 5C). Transfection of Luc_58–165Z did not result in increased expression of Luc10UGA in the media or lysate while Z_58–165Luc showed robust expression of the protein in the lysate (Fig 5B). qRT-PCR analysis confirmed that the chimeras had similar RNA levels (Fig 5C). Fig 5D shows the efficiency of full-length ZSELENOP expresion after normalizing for mRNAs levels and transfection efficiency. The Z_58–165Luc chimera resulted in an almost 5-fold increase in efficiency compared to the Δ58–165 mutant, suggesting that the distance between the first UGA and the signal peptide is crucial for efficiency of Sec incorporation. It is possible that 100% recovery of the Z_58–165Luc chimera requires the native sequence. These results suggests that the 58–165 sequence acts as a spacer element between the first UGA and the start codon and/or signal peptide.

FIGURE 5. Distance between the first UGA and the consensus signal peptide influences Sec incorporation effficiency.

A) Schematic diagram of the SELENOP chimeras for the N-terminal deletion mutant Δ58–165. B) Top and Bottom panel ⁷⁵Se labeled media and lysate respectively from transient cell lines expressing chimeras shown in A, resolved by SDS-PAGE and detected by PhosporImager analysis. The arrows indicate full-length and termination product. C) qRT-PCR analysis of mRNA data obtained for samples from B as described in Fig 1B. D) Efficiency expressed as a ratio of protein band intensity in media and lysate in B to fold change in RNA for the ZSELENOP N-terminal chimera relative to wild type.

Both SELENOP N-terminus coding sequence and UGA codon context can independently drive multiple Sec incorporation efficiency.

The N-terminal coding sequence is able to stimulate Sec incorporation of downstream UGAs placed in an non-native context both in cells and in vitro (Fig 3A). (35). In our study the first UGA is located in the N-terminus and we therefore asked if the assembly of the Sec incorporation complex at the first UGA in its native context is required to stimulate Sec incorporation at downstream UGAs placed in a non-native context. We therefore created variants of Luc10UGA, ZN-term_Luc and Luc_ZC-term in which the first Sec (U) at amino acid position 59 was mutated to a Cys (C) and termed them as Luc10UGA_U59C, ZN-term_Luc_U59C and Luc_ZC-term_U59C. HEK293 cells were transiently transfected with wild-type ZSELENOP, Luc10UGA, ZN-term_Luc and Luc_ZC-term and their U59C variants. Both media and lysate from transfected cells were analyzed for expression using ⁷⁵Se labeling.

As shown in Fig 6A-B, transient transfection of all three U59C mutants of Luc10UGA, ZN-term_Luc and Luc_ZC-term resulted in an increased efficiency of full length protein production. The U59C mutants of both N- and C-terminal chimeras as shown in Fig 6A resulted in either similar or higher levels of full length product relative to wild-type ZSELENOP. Both ZN-term_Luc and its variant ZN-term_Luc_U59C were able to stimulate Sec incorporation efficiency of downstream UGAs that were placed in an non-native context. Interestingly, the almost completely defective Luc_ZC-term is effectively rescued by its U59C variant. This suggests that in the absence of the N-terminus and the first UGA, the C-terminal sequence context can also support processive and efficient Sec incorporation thus supporting the role for UGA codon context. Together these data indicate that the SELENOP N-terminus sequence contains efficiency elements which is active even when it does not incorporate/contain the first UGA and in the absence of the N-terminus sequence the C-terminal sequence context is “programmed” to allow processive Sec incorporation at downstream UGAs. Thus we have identified two independent mechanisms of regulation for efficiency of processive Sec incorporation.

FIGURE 6. SELENOP N-terminus remotely drives multiple Sec incorporation efficiency.

A) Top and Bottom panel ⁷⁵Se labeled media and lysate respectively from transient cell lines expressing U59C mutants of chimeras shown in 3A, resolved by SDS-PAGE and detected by PhosporImager analysis. The arrows indicate full-length product and endogenous selenoproteins. B) Percent efficiency of full length product protein band intensity in media and lysate normalized for transfection efficiency relative to wild type. Data is plotted as the average plus and minus standard deviation for three independent experiments. P < 0.05 compared to value with the native ZSELENOP using paired two-tail students t-test. C) Top and Bottom panel ⁷⁵Se labeled media and lysate respectively from transient cell lines expressing N- and C-terminal chimeras shown in 3A and Luc10UGA with its first UGA in its native context spanning 150nts in the presence of zebrafish signal peptide

DISCUSSION

To answer the key question whether SELENOP synthesis employs novel RNA elements or cellular factors to accommodate processive and efficient Sec incorporation, we optimized the zebrafish SELENOP expression using C-terminal FLAG tag and ⁷⁵Se labeling in HEK293 cells. A zebrafish SELENOP system has been used previously to characterize the SELENOP 3’UTR but the construct contained an N-terminal GST tag and a significant amount of termination at the second UGA was observed (32). Using the native zebrafish cDNA, we were not only able to detect robust expression, but we also found that most of the product was full length protein. Surprisingly, we were unable to express transfected human SELENOP cDNA in HEK293 cells. There is one report of successful expression of mammalian SELENOP in mammalian cells (41), but we have not been able to reproduce that result with rat, human or mouse SELENOP. It is interesting to note that the expression of either the wild-type version of human SELENOP or its all-Cys variant alter the endogenous SELENOP synthesis that favors the early termination product. We interpret the sum of these results as further evidence for tight regulation post-transcriptionally and perhaps limited translational capacity. The fact that ZSELENOP mRNA can be translated to generate full length protein in mammalian cells indicates strong conservation of the core mechanism required for processive Sec incorporation. Thus, in this study we have exploited the use of this system to begin teasing out the key components of the SELENOP mRNA that contribute to these regulatory events (summarized in Fig 7).

FIGURE 7.

Schematic of constructs used in the study and their expression relative to wild-type ZSELENOP.

In vitro translation of mRNA encoding an artificial selenoprotein with multiple UGA codons (Luc10UGA) results in a highly inefficient synthesis of full length product (35). Substituting the N- or the C-terminus of Luc10UGA with either the N- or C- termini of native sequences from human SELENOP rescues efficiency by 40 and 60%, respectively in vitro (35). This suggests a role for the native SELENOP sequence and more specifically UGA codon context. This finding is congruous with other studies where the UGA codon context has been shown to dictate efficiency (19, 42). Our analysis of the ZSELENOP in cells indicates that the N-terminal portion of SELENOP is sufficient to drive downstream Sec incorporation processively irrespective of whether it contains the first UGA. This is especially interesting since both the N-terminal chimeras (ZN-term_Luc and ZN-term_Luc_U59C) have all their downstream UGAs in a non-native context, thus overruling the role for UGA codon context and indicating that the N-terminal sequence has the dominant ability to dictate downstream Sec incorporation even when it does not contain its own Sec codon.

Our systematic scan of the ZSELENOP coding region for regulatory elements revealed that only Δ58–165 had a significant on efficiency (5–10-fold) while its cysteine version could be robustly expressed. However, the Δ58–165 mutant was rescued with a non-specific sequence from the artificial selenoprotein, thus confirming the role for the 58–165 nts as a spacer element and not an efficiency sequence. The reduced expression from the Δ58–165 deletion mutant is likely the result of increased proximity between the first UGA and the start codon or the signal peptide. The minimal distance between the start codon and an in-frame UGA required for maximal Sec incorporation efficiency has been reported previously as 21 nts for GPX1 (20). In native SELENOP with the signal peptide, the distance between the start codon and the first UGA is 174 nts which is reduced to 66 nts in the Δ58–165 mutant, thus maintaining the minimal distance between the start and the Sec codon that is required for efficient Sec incorporation. On the other hand, the distance between the first UGA and the signal peptide in the native sequence is 117nts which is reduced to 9nts in the Δ58–165 mutant. Thus, the proximity of the first UGA to the signal peptide in the Δ58–165 mutant most probably results in interference between the Sec incorporation machinery and signal peptide recognition complex thereby resulting in reduced efficiency. Moreover, the signal peptide in zebrafish has not yet been experimentally defined and may extend longer than the known consensus sequence.

Surprisingly, we were unable to identify any discrete N-terminal sequence that dramatically disrupted the overall efficiency of full length product (Fig 4E). However, three of the five N-terminal deletions resulted in multiple apparent early termination products (Fig 4A). We speculated if this effect was due to the deletion mutant resulting in reduced distance between the first UGA and the C-terminal Sec codons. But this may not be true since Δ616–765 mutant did not show any major processivity defects. Thus we propose that the N-terminal sequences may be redundant in function for processivity.

In contrast to in vitro studies, the C-terminal chimera (Luc_ZC-term) in cells, is defective for both efficiency and processivity (Fig 3B). Strikingly, this defective C-terminal chimera (Luc_ZC-term) is fully rescued when the first Sec codon (U59) is mutated to Cys, thus endorsing the importance of the C-terminal sequence context. Thus the C-terminal sequence can also independently drive efficient and processive Sec incorporation in the absence of the first UGA. The relevance of the the C-terminus sequence is further supported by the result that the N-terminal sequence is able to independently regulate processive Sec incorporation but with only 54% efficiency for full length product compared to wild type. Furthermore we find that one of the deletion mutants in the C-terminal sequence Δ766–915 showed a 2-fold reduction in efficiency (Fig 4E), despite the reduced number of Sec codons. Stem loop structures termed as SRE2 and SRE3 have been previously bioinformaticaly predicted in the C-terminus of SELENOP but their function remains unknown (35). It is possible that the predicted SRE2 and SRE3 may dictate the processive Sec incorporation machinery in the C-terminus .

Given the 100% conserved position of the first UGA at nt 59–61 in all vertebrates and the observation of significant termination at this position, it has been predicted to be the bottleneck for downstream Sec incorporation (32). To analyze the role for the first UGA in processive Sec incorporation and efficiency, we tested both a deletion mutant and also a U59C mutant. The 166–315 sequence contains the first UGA and also the bioinformatically identified SRE1 structure (35). Both the deletion mutant Δ166–315 and the U59C mutant increased the Sec incorporation efficiency by ~20%. This increase was more dramatic in the U59C mutants of the the N-and C-terminus chimeras. These results corroborate earlier studies predicting slow decoding efficiency at the first UGA position. However, the fact that the chimera lacking the N-terminus does not generate full length product unless the first UGA is mutated to Cys indicates that the first UGA must be in a specific context.

In this study we have identified two key findings. First, the N-terminal sequence can remotely stimulate Sec incorporation of downstream UGA codons placed in a non-native context. Second, in the absence of the first UGA, the C-terminal sequence is sufficient to drive efficiency and processivity, suggesting a role for UGA codon context. Thus we have identified a dual mode of regulation within the SELENOP coding region sequence. This then raises the question of why is there a need for the two seemingly independent mechanisms. If UGA codon context is sufficient then why is there a need for the regulation by the N-terminus sequence. Typically in cells and in plasma samples, two main isoforms of SELENOP has been identified (37, 43). One is the full length protein and a second that terminates at the 2nd UGA. The ratio of the full length protein to early termination has been reported to be affected by polymorphisms both in the coding region and in the 3’UTR and also by selenium supplementation. The need to maintain the relative abundance of full length and truncated protein that terminates at the 2nd UGA is not completely understood. One study showed that the 2nd UGA termination isoform is an active N-terminal domain which contains a thioredoxin-like redox active site and functions as an antioxidant (44). Plasma samples from a selenium deficient population contain a larger ratio of early termination isoform at 2nd UGA to full length product (36). This enzymatically active SELENOP isoform may be crucial to maintain cellular redox homeostasis during selenium deprived conditions, and its synthesis may be specifically regulated. In addition, synthesis of the smaller SELENOP isoform seems logical during selenium limiting conditions as this would make selenium available for expression of other essential selenoproteins such as GPX4.

We propose that in cells with adequate selenium, the SELENOP N-terminus adopts a structure that favors the recruitment of the processivity complex and/or efficiency factor/along with the Sec incorporation complex and this complex stably associates with the translating ribosome all the way up to the natural stop codon. The activity of the N-terminus is enhanced by the C-terminal sequence context. Meanwhile in limiting selenium levels, the N-terminus undergoes structural rearrangement that engages a Sec incorporation complex that is devoid of the processivity/efficiency factor(s). This inhibits the Sec incorporation complex from efficiently proceeding to the downstream UGAs and thus mostly results in 2nd UGA truncated isoform.

Currently there is a dearth of information on the structure and factors specifically binding SELENOP mRNA. Future studies analyzing the structure and nature of ribonucleoprotein complexes on SELENOP mRNA should provide new insight into the mechanism of the regulatory elements identified in this study.

CONCLUSIONS

This study provides new insight into the mechanism of processive Sec incorporation at a sequence level. How these sequences contribute independently as structural elements or associate with trans factors for function remains to be determined. This study lays the groundwork for future analysis and identification of ribonucleoprotein complexes that associate with endogenous SELENOP mRNA.

MATERIALS AND METHODS:

Constructs-

The coding region and full 3’ UTR of zebrafish SELENOP (ZSELENOP) in pcI-Neo vector was a gift from Dr. Vadim Gladyshev (Harvard Medical School, Boston, MA). This construct was FLAG-tagged (GATTATAAGGATGATGATGATAAG) on the C-terminus using the QuikChange Site-Directed Mutagenesis Kit (Agilent) as per manufacturer’s instructions, and the resulting product was TA cloned into a mammalian expression vector, pcDNA3.1 using TOPO - TA cloning kit (Invitrogen). All deletion mutants were generated using the QuikChange Site-Directed Mutagenesis Kit (Agilent) as per manufacturer’s instructions. The resulting mutant sequences were confirmed by DNA sequence analysis. For the coding region of the artificial selenoprotein (Luc10UGA), the N-and C-terminus chimeras (ZN-term_Luc and Luc_ZC-term) were synthesized commercially from Integrated DNA technologies (IDT) as described earlier (35). Briefly, to design the artificial Luc10UGA we used the luciferase coding region, which was truncated to match the length of the human SELENOP mRNA, and the Sec codons were positioned in this truncated construct in exactly the same position as found in human SELENOP. Similarly, the mutants in which the the 58–165 nts sequence was interchanged between Luc10UGA and wild type zebrafish (Z_58–165 Luc and Luc_58–165_Z) were synthesized commercially from integrated DNA technologies. All these constructs were then cloned into pcDNA3.1 vector and the ZSELENOP 3′UTR was then ligated into the Sec containing constructs using Pac1 and Not1 linkers. To create the U59C mutants of Luc10UGA and the N-and C-terminus chimeras (ZN-term_Luc and Luc_ZC-term) we used QuikChange Site-Directed Mutagenesis Kit (Agilent) as per manufacturer’s instructions. The resulting sequences were confirmed by DNA sequence analysis.

Cell culture and transfection—

HEK293 cells (ATCC# CRL1573) cultured in Hyclone EMEM media containing 10% (v/v) heat inactivated fetal calf serum (FBS) and maintained at 37°C with a humidified 5% CO2 atmosphere were used for transient and stable transfection of wild type and deletion mutants of ZSELENOP. For transfection, jetPRIME (Polyplus) reagent was used according to the manufacturer’s protocol. Cell density of 5–6 ×10⁵ per well were seeded 24 hours prior to transfection in a 6-well plate. After 48 hrs of transfection, media was supplemented with G418 for selection of stables. Stably selected cells were maintained in EMEM media with 0.63 µg/ml of G418. For transient transfection, cells were seeded at a density of 8 ×10⁵ in six well plates and transfected the next day.

⁷⁵Se labeling-

Stably transfected cell lines were seeded at a density of 8 ×10⁵ in six-well plates and 24 hours later the media was changed to a serum free EMEM media supplemented with 100 nM of ⁷⁵Se (specific activity 6.29µCi/µl; Research Reactor Center, University of Missouri, Columbia). Similarly, for transiently transfected cells, 24 hours post transfection the media was changed to a serum free EMEM media which was supplemented with 100 nM of ⁷⁵Se. After ~16 hours, the media was collected and centrifuged at 2500 x g for 5 minutes at 4°C. The top 80% of the centrifuged media was transferred into a new tube. 1.5% of the total centrifuged media was used for analysis by 12% SDS-PAGE. The adhered cells were then gently washed with cold PBS and lysed in 1% NP-40 buffer (50mM Tris-HCl pH8.0, 150mM sodium chloride, 1% NP-40, Roche complete Protease Inhibitor). The lysate was then cleared by centrifugation at 17,000 x g for 10 min at 4°C. 20% of clarified lysate was used for protein analysis and the remaining lysate was used for total RNA purification. Total protein concentration was analyzed by using detergent compatible BCA protein assay (Pierce). Equal protein amounts of both media and lysate were resolved by 12% SDS-PAGE and visualized by PhosphorImaging (GE Healthcare) or by 12% SDS-PAGE for western blot analysis.

Western blot-

15 µl of the lysate (10% of the total cell lysate which is ~30 ug) as described above was then resolved by 12% SDS-PAGE blotted to membrane (Amersham Biosciences), blocked overnight in 5% nonfat dried milk for 1 hour at room temperature, and probed using a monoclonal horseradish peroxidase-conjugated (HRP) anti-FLAG antibody (Sigma) at a 1:5,000 dilution overnight, and was developed using the SuperSignal West Femto kit (Pierce) according to the manufacturer’s protocol.

qPCR- cDNA synthesis-

Total RNA was purified using the RNAeasy kit (Qiagen). Total RNA was reconstituted using nuclease free water and quantitated using a Nanodrop 2000c spectrophoptometer. 200 ngs of total RNA was converted to cDNA using the reverse transcriptase (RT) Superscriptase III (Applied Biosystems). 20 µl reactions were set up according to the manufacturer’s protocol using oligo dT primer.

FLAG Immunoprecipitation-

Magnetic beads bound with anti FLAG antibody were equilibrated with buffer A (0.2 M Tris pH 7.5, 0.2 M KCl, 0.001 M EDTA pH 8.0, 0.5 M NaCl, 25% glycerol and 0.01 M Tween) and then incubated with 75Se labeled media from stable transfectants of either wild type or mutants of ZSELENOP for 4 hours. Beads were then washed with Buffer A and then Buffer B (0.2M Tris pH 7.5, 0.2M KCl, 0.001M EDTA pH 8.0). Bound protein was eluted using SDS PAGE buffer or 110ug/ml of 3XFLAG peptide in 1ml Buffer B.

qRT-PCR and data analysis--

Gene specific forward and reverse primers were designed to amplify 57–71 base long reference and target amplicons. Actin was used as reference gene. Forward and reverse primers used to amplify HSELENOP were CTAGGAGCTGATGCTGCCATT and GGTGATTGCAGACCCTGTTTTT respectively. Similarly, forward and reverse primers for ZSELENOP 3’UTR were TGCTGGTTAACACTCAAGGTGAA and AAGTCCAGTGCTGCCTCACA and for coding region were AACAGAAGGAGCCCGCTGTAA and GTTCTTCATGGGCTCCACATC primers. Actin forward and reverse primers were GCGCGGCTACAGCTTCA and CTTAATGTCACGCACGATTTCC respectively. qRT-PCR was performed using an ABI Step One Plus Real Time PCR System (Applied Biosystems) and PCR Master Mix (ABI) with Power SYBR Green (Invitrogen) and ROX reference Dye (Invitrogen). All reactions were performed in triplicate. The total reaction volume was 20 µl with 5 µl of 1:5 diluted RT reaction. Working concentration of the primers in the reaction was 0.25 µM. Thermal cycling conditions were 95°C for 10 min followed by 40 cycles of 95°C for 15 sec, 60°C for 1min. Melt curve analysis was performed for each sample to ensure a single amplification product. Samples were analyzed in triplicate for both the reference gene and the target gene. Quantitation was performed using the comparative ΔΔCt method. We used actin as the normalizer and in Figure 1, the calibrator sample used was the endogenous human SELENOP from untransfected samples since it did not significantly change upon transfection with ZSELENOP. In all the other figures, transfected wild type ZSELENOP served as the calibrator sample. Primers in the acceptable efficiency range (90–110%) were determined using the standard curve method.

HIGHLIGHTS.

• Full length SELENOP synthesis requires processive selenocysteine (Sec) incorporation at multiple in-frame UGA codons in the coding region.

• Both SELENOP N-terminal and C-terminal sequences can independently drive processive and efficient Sec incorporation at downstream UGA codons.

• The distance between the first UGA and the signal peptide is critical for efficiency of full length SELENOP synthesis.

• We have identified coding region sequences within the SELENOP mRNA that influence Sec incorporation efficiency and processivity at multiple UGA’s.