Impairment of alternative splice sites defining a novel gammaretroviral exon within gag modifies the oncogenic properties of Akv murine leukemia virus

Background Mutations of an alternative splice donor site located within the gag region has previously been shown to broaden the pathogenic potential of the T-lymphomagenic gammaretrovirus Moloney murine leukemia virus, while the equivalent mutations in the erythroleukemia inducing Friend murine leukemia virus seem to have no influence on the disease-inducing potential of this virus. In the present study we investigate the splice pattern as well as the possible effects of mutating the alternative splice sites on the oncogenic properties of the B-lymphomagenic Akv murine leukemia virus. Results By exon-trapping procedures we have identified a novel gammaretroviral exon, resulting from usage of alternative splice acceptor (SA') and splice donor (SD') sites located in the capsid region of gag of the B-cell lymphomagenic Akv murine leukemia virus. To analyze possible effects in vivo of this novel exon, three different alternative splice site mutant viruses, mutated in either the SA', in the SD', or in both sites, respectively, were constructed and injected into newborn inbred NMRI mice. Most of the infected mice (about 90%) developed hematopoietic neoplasms within 250 days, and histological examination of the tumors showed that the introduced synonymous gag mutations have a significant influence on the phenotype of the induced tumors, changing the distribution of the different types as well as generating tumors of additional specificities such as de novo diffuse large B cell lymphoma (DLBCL) and histiocytic sarcoma. Interestingly, a broader spectrum of diagnoses was made from the two single splice-site mutants than from as well the wild-type as the double splice-site mutant. Both single- and double-spliced transcripts are produced in vivo using the SA' and/or the SD' sites, but the mechanisms underlying the observed effects on oncogenesis remain to be clarified. Likewise, analyses of provirus integration sites in tumor tissues, which identified 111 novel RISs (retroviral integration sites) and 35 novel CISs (common integration sites), did not clearly point to specific target genes or pathways to be associated with specific tumor diagnoses or individual viral mutants. Conclusion We present here the first example of a doubly spliced transcript within the group of gammaretroviruses, and we show that mutation of the alternative splice sites that define this novel RNA product change the oncogenic potential of Akv murine leukemia virus.


Background
Many murine leukemia viruses (MLVs) belonging to the genus gammaretroviruses induce cancer when injected into susceptible newborn mice [1,2]. These simple retroviruses do not themselves harbor transduced oncogenes, and their ability to cause cancer relies on the host cellular genes that are transcriptionally activated or otherwise mutated as a result of the integrated provirus [3][4][5][6].
Regarding the virus itself, it is well documented that the LTR region plays a crucial role for both the strength and cell type specificity of disease induction [7,8]. Within the LTR the specificity has been located mainly to the enhancer region in U3, and further narrowed down to the sequences defining different transcription factor binding sites [9][10][11][12]. In spite of this predominant role of the LTR in MLV pathogenesis, also sequences outside this region have been shown to be important for the ability and potency of a particular virus to induce cancer. Infection is mediated by interaction between the viral envelope protein (Env) and a specific host cell receptor, and for the ecotropic MLVs such as Moloney, Akv, and SL3-3, this receptor has been identified as the mouse cationic amino acid transporter 1 (mCAT1) [13,14]. A significant role of env in MLV pathogenesis is the involvement in the generation of recombinant polytropic viruses that takes place during T-cell lymphoma development. These MCF (mink cell focus-forming) viruses have the ability to superinfect cells, an aspect which is thought to contribute to tumor formation [15,16]. In addition to the env gene, and perhaps somewhat surprisingly, the viral gag gene sequences have also proven to play a role in MLV pathogenesis. Thus, Audit et al. (1999) [17] showed that the introduction of only three synonymous nucleotide mutations in the capsid-coding gene of Moloney MLV (Mo-MLV) changed the oncogenic properties of this virus. The mutations were located at an alternative splice donor site (SD'), which together with the canonical env splice acceptor site was shown to produce a subgenomic transcript of 4.4 kb [18]. The equivalent transcript, produced by Friend MLV, was subsequently shown to be packaged into virions, reversely transcribed and integrated in the host genome by normal viral mechanisms [19]. While wild-type Mo-MLV induces T-cell lymphomas in 100% of the inoculated mice, the SD' mutant virus exhibited a much broader specificity, thus inducing -besides the expected T-cell tumors -erythroid or myelomonocytic leukemias. In contrast, the corresponding mutations in a Friend MLV background did not seem to influence the pathogenic potential of this virus at all. Both wild-type and mutant Friend MLVs induced exclusively the characteristic erythroleukemia [17]. So it seems that the importance for the diseaseinducing potential of the SD' site, although conserved among many species, is strongly dependent on the virus type.
The SD' site has also been found to be used for production of the oncogenic gag-myb fusion RNAs in promonocytic leukemias induced by Mo-MLV in pristane-treated BALB/ c mice [20]. When the SD' site was mutated in this model, the overall disease incidence was not affected; however the proportion of myeloid leukemia decreased significantly, while the proportion of lymphoid leukemia increased. Moreover, no 5' insertional activation of c-myb (using alternative splice donor sites) could be found, thereby signifying a specific requirement of the SD' site for this mechanism [21].
Here we report of the identification of an alternative splice acceptor site, SA', located in the capsid region of gag, which together with the gag splice donor site, SD' (corresponding to the one reported for Moloney and Friend MLV), or together with a second alternative gag splice donor site, SD*, defines a novel exon within the genus gammaretroviruses. We show that RNA splicing by use of the alternative splice sites does indeed take place in tumor tissue, and that both double-and single-spliced transcripts are produced. When mutating the SD', the SA', or both sites simultaneously, the splicing pattern is affected in a predictable way. Moreover, we demonstrate that the SA' and SD' mutations alter the oncogenic specificity of the Akv MLV, displayed by a change in the distribution of the diagnoses of the resulting tumors as well as by an induction of tumors of altered specificity such as histiocytic sarcoma and de novo diffuse large B cell lymphoma (DLBCL).

Identification of a novel exon residing within the gag region of Akv MLV
In order to identify potential alternative splice donor and splice acceptor sites in Akv MLV, exon-trapping was performed using the exon-trapping vector pSPL3 (see Materials and Methods). In short, an exon resulting from usage of the alternative splice acceptor (SA') and either one of two alternative splice donor (SD' or SD*) sites located in the capsid region of gag (Fig. 1), was isolated and verified by RT-PCR analyses of RNA isolated from Akv MLV infected cells (data not shown). The size of the exon is 235 bp or 180 bp, depending on the splice donor site used.

Mutations of the alternative splice sites affect the specificity of the induced tumors
To analyze a possible effect in vivo of the novel exon, defined by SA' and SD', three different alternative splice site mutant viruses, Akv-CD, Akv-EH, and Akv-CDH, mutated in either the SA' or SD' site, or in both sites simultaneously, were constructed and injected into newborn mice of the inbred NMRI strain. Fig. 1 shows the precise locations of the synonymous mutations around the trapped exon. Without altering the coding potential of the capsid gene, the mutations affect the branch point site, the pyrimidine region, the conserved splice junction AG and GT dinucleotides, and the fairly well-conserved exonal A at the SD' junction site. The positions of the three intron mutations at the SD' junction site are identical to those in Moloney and Friend MLV described by Audit et al. (1999) [17].
As can be seen from Fig. 2 Fig. 3) and diagnosis according to the Bethesda classification [22] revealed that a large proportion (approx. 70%) of the total numbers of tumors could be classified as either follicular B-cell lymphoma (FBL) (13%), diffuse large B-cell lymphoma (DLBCL) progressed from FBL (33%), or plasmacytoma (PCT) (25%) ( Table 2). However, the distribution was quite different within the different virus series; thus, almost one quarter of the Akv-wt induced tumors were diagnosed as FBL, while no tumor of the Akv-CD group (p < 0.05) or one tumor each of the Akv-EH or Akv-CDH groups fell into this group. In contrast, within the DLBCL tumors progressed from FBL the frequencies are similar (ranging from 24% to 39%) no matter if the causative virus contained mutated SA' and/or SD' sites or not. In the PCT group it appears that mutating the SA' site significantly impaired the ability of the virus to induce PCT (p < 0.05). On the other hand, this effect was not statistically significant if the SD' site was mutated, and curiously if both sites were mutated, wild-type level for PCT induction was restored.
In line with this, the most dramatic effect in general was seen when only the SA' site was mutated as shown for Akv-CD; the tumor incidence of this mutant with respect to splenic marginal zone lymphoma (SMZL) increased from Location of the trapped exon  8% to 28% (p < 0.1) and decreased to 0% as shown for Akv-EH (p < 0.05) and for Akv-CDH (p = 0.5). Moreover, the Akv-CD mutant virus was the only one that displayed a capability for inducing histiocytic sarcoma, a tumor type which has not been observed in any of our previous studies using NMRI mice (inbred or random-bred) infected with Akv, SL3-3, or different derived mutants of these. So in brief, synonymous mutations at the SA' site of Akv MLV markedly altered the oncogenic potential of the virus by significantly impairing the ability to induce both FBL and PCT. Besides, while the development of SMZL was increased by Akv-CD, it was abolished in Akv-EH and Akv-CDH, and most notably, a novel potential for inducing histiocytic sarcoma was established.
The most pronounced effect of mutating the SD' site (Akv-EH) is the frequent occurrence (35%) of diffuse tumors, which according to the Bethesda classification represent DLBCL centroblastic (more than 50% of the infiltrating population is centroblasts). These tumors, where progression is not from either a follicular or a marginal lym-phoma, are comparable to the de novo lymphomas in humans, and to emphasize this association we have used the term de novo DLBCL (Table 2). Strikingly, de novo DLB-CLs were never observed among the wild-type induced tumors or among the other mutant induced tumors (p < 0.05). The finding of such tumors in mice is rare and could be exploited to understand the molecular changes in de novo DLBCL of mice, and eventually a useful mouse model of human de novo DLBCL might be generated from this set-up.
Quite unexpectedly, the effect of mutating the SA' and SD' sites simultaneously (Akv-CDH) was the less manifested one. FBL incidence dropped from 23% to 7%; otherwise this mutant in our experimental setting displayed similar tumorigenic potential as the wild-type Akv MLV.

Conservation of the introduced splice site mutations in the tumors
To determine the stability of the introduced mutations, the regions containing the mutations were PCR amplified Pathogenicity of Akv and derived splice site mutants in inbred NMRI mice   from genomic DNA prepared from the induced tumors, using the primers depicted in Fig. 1. The sequences of the amplified fragments confirmed in all cases the integrity of the introduced mutations (data not shown).

Both single-and double-spliced transcripts are generated in vivo
The observed effect of the mutated splice sites on the oncogenic properties advocates that RNA splicing by means of the alternative SA' and SD' sites does indeed take place in vivo. To clarify and confirm the identity of the produced transcripts, the splice pattern in tumor tissues (and for comparison in NIH 3T3 cells infected with the same four viruses) was analyzed. RNA from the individual endstage tumors (or from virally infected cells) was isolated, and conventional RT-PCRs were performed with primers designed in such a way that it should be possible to identify all four potential splice products using 4 different primer sets as shown in Fig. 4A.
With a few exceptions, all tumors were analyzed, and sequences of the amplified RT-PCR products determined to validate the specificity of the fragments (data not shown). Representative results from each virus series are shown in Fig. 4B. In all cases, PCR products representing splice product A (the regular env transcript; primer set #4) was observed, which implies that damage of the alternative splice sites, SA' and SD', does not impair the production of the regular single-spliced env RNA. Concerning splice product D (primer set #1) it was never amplified, neither from tumor tissues nor from cell culture studies, strongly indicating that this is not a bona fide transcript.
The lack of detection of product D is unlikely to result from a technical PCR-problem since the two primers have been validated in other PCRs.
For the Akv wild-type induced tumors, RT-PCR products representing the double-spliced product B (primer set #2), and fragments of expected size amplified by primer set #3, indicative of splice product B or C, were observed in all cases. As would be expected primer set #2, which is dependent on an intact SA' site, did not result in any amplification products using RNA from Akv-CD or Akv-CDH tumors. Surprisingly however, in five out of 14 analyzed Akv-EH tumor samples (represented by Akv-EH tumor no. 14 in Fig. 4B), a product slightly smaller than that of transcript version B was amplified. The subsequent sequence analyses revealed that the alternative splice donor site SD* (depicted in Fig. 1) in these cases consistently had been used, resulting in the generation of a splice product equivalent in structure to product B, however 54 nucleotides shorter. No correlation between tumor cell specificity and usage of the SD* site could be observed, since the five tumor samples originated from FBL, DLBCL progressed from FBL, de novo DLBCL, and the single case of STL (small T-cell lymphoma). The presence of the same splice product from the SD* site was verified by sequence analysis of RT-PCR products derived from tumors induced by the wild-type virus in some cases, although the product was consistently less prominent than product B.
Transcript C corresponds to the single-spliced transcript of 4.4 kb, which previously has been reported to be produced by both Friend and Moloney MLV using the SD' together with the canonical env SA' site [18,19]. Our RT-PCR results confirm the existence of this single-spliced transcript, since products of the expected size were always amplified with primer set #3 using RNA from Akv-CD tumors (Fig. 4B), whereas product B (primer set #2) was never amplified in this material.
In summary, by means of the alternative splice sites that define the novel gag exon, both a single-spliced transcript C as well as a novel double-spliced transcript B is produced in vivo, and when these alternative splice sites are destroyed, the splicing pattern is changed concordantly. The same RT-PCR analyses were performed for NIH 3T3 cells infected with the four viruses, which led to the same splice pattern (data not shown). In addition, Northern blot hybridizations with an ecotropic env-probe and with a probe covering the novel SA'-SD' defined exon in gag were performed with RNA isolated from these cells (Fig.  5). Besides the expected hybridization patterns of prominent bands of full-length (env and gag probe) and env mRNA (only env probe) sizes, a weaker band of a size corresponding to splice product C (4.4 kb) was detected with both probes. No distinct band corresponding to spliced RNA B was observed, suggesting a very low level of production and/or significant messenger instability.

Provirus integration site analyses
In order to identify a possible connection between specific retroviral integration sites (RIS) and specific diagnostic tumor types, provirus integration sites from the majority of the induced tumors were isolated and sequenced. We have then by subsequent homology searches of the mouse genome databases identified 240 unambiguous integration sites (Table 3). These integration site sequences rep-  [24,25], both version mm8, 111 novel RISs were identified. In an attempt to pick up candidate cancer genes that might be associated with specific tumor diagnoses, we looked for common integration sites (CISs), which would infer such genes [25,26]. Hence, we compared the Northern blot hybridizations with an ecotropic specific env probe and a gag probe of RNA isolated from NIH 3T3 cells chron-ically infected with the viruses listed above each lane      The gene (or RefSeq) closest to the integrated provirus is given (UCSC, mouse mm8 assembly). ---indicates that the distance to the closest gene/RefSeq is more than 100 kb. b For each insertion, it is indicated, based on RTCGD (mm8), whether a novel RIS has been defined. c For each insertion, it is indicated, based on RTCGD (mm8), whether a novel CIS has been defined. The definition follows the recommendations from RTCGD with a window size of 100 kb, 50 kb, and 30 kb for CISs with 4 (or more), 3, or 2 insertions, respectively. * indicates an exception from this rule, if two integration sites are found within the same gene/RefSeq. n.d., not determined.  integration sites with each other as well as with previously defined RISs in RTCGD. In principle, using the recommendations from RTCGD with a window size of 100 kb, 50 kb, and 30 kb for CISs with 4 (or more), 3, or 2 insertions, respectively, we were hereby able to define 35 novel common integration sites (CISs) ( Table 3). Just a single one of these could be correlated with a specific diagnosis and with a specific virus, since in two independently Akvwt induced plasmacytomas a definite region of chromosome 15 was targeted. However, this region does not contain genes/RefSeqs within a 100 kb distance from the integrated proviruses, so for the present we cannot predict what -if any -candidate gene(s) that might have been influenced by the integrated proviruses. For the remaining 34 CISs more than one virus and more than one tumor diagnosis were implicated, which implies that no straightforward association between target gene (and/or causative virus) and tumor type can be deduced.

kb (spliced env)
In six cases, the same chromosomal locus was targeted several times. These cases include Bach2 (hit 4 times), Sox4 (hit 4 times), Hhex (hit 5 times), Ii (hit 5 times), a region of chr. 16 not containing any genes/RefSeqs within a distance of 100 kb from the integration sites (hit 5 times), and LOC433632 (hit 7 times). Five of these integration sites were already registered in RTCGD, only the Ii locus define a completely novel RIS/CIS. This latter finding may suggest that Ii targeting is strongly associated with the applied model system (virus/mouse strain) [27]. We also note that an integration has taken place within the first intron of Stk10 in a plasmacytoma induced by Akv-wt (Table 3, #140), which appears to be in conflict with the work of Shin et al., 2004 [28] where Stk10 was described as a SMZL specific candidate gene.
Finally, we examined if specific regions of the targeted gene/RefSeq have been favored with respect to orientation and position of the integrated provirus. We have recently reported of differences between Akv MLV and an enhancer mutant hereof, Akv1-99, in their patterns of proviral insertions around host transcription units in the induced tumors [29]. In line with this, it might be envisioned that destroying the alternative splice sites of the virus could lead to a different pattern of integration site selection during tumorigenesis; e.g. it might be speculated if particular positions relative to the target gene somehow would facilitate gene deregulation dependent on the presence or absence of intact SA' and/or SD' sites. Accordingly, we allocated each individual integration site position and orientation to a defined region, i.e. either upstream, within the promoter, within 1. intron, within last intron, within all other introns, within exons, or downstream of the target gene (Table 4). Eleven cases in total were excluded as they were positioned -almost with the same distance -in between two target genes, upstream of one and down-stream of the other. As seen in table 4, no clear differences with respect to the four viruses were observed, signifying that mutations of the alternative splice sites do not have major effect on the ability of the Akv MLV to affect the target gene/RefSeq from certain positions and/or orientations. It might be worth to notice that the overall picture shows that about half of all integrations are found within introns, and among these there appear to be a tendency for a provirus orientation opposite to that of the target gene. In these cases the formation of chimeric RNA species by promoter insertion and/or splicing would not be predicted.

Discussion
We have in the B-lymphomagenic Akv MLV identified a novel exon, which is defined by the alternative splice acceptor (SA') and the splice donor (SD') sites located in the capsid encoding region. While previous studies of Moloney and Friend MLV have demonstrated production of a 4.4 kb transcript using the same SD' site together with the canonical env SA site, this is the first report demonstrating the existence of an alternative SA' site and production of a double-spliced transcript during the life cycle of a replication-competent simple retrovirus. Yet it remains to be investigated how widespread this competence is. An alignment between six murine retroviruses shows that the conserved splice junction dinucleotide AG is present neither in Cas-Br-E nor in Moloney MLV, although the region in general is well-conserved (Fig. 6).
We did not perform detailed analyses of the influence of the splice site mutations on the viral replication. However, since the same number (10 5 to 10 6 ) of infectious virus particles, as measured by infectious center assays, were injected from each virus series, and since the mutant viruses induced tumors with comparable incidences and latencies as the wild-type virus, it is not likely that the mutations had imposed severe weakening on the in vivo spreading capability. Hence the observed shifts in specificity of the induced tumors most likely are a direct result of the introduced mutations. However, we note that Houzet The in vivo significance of the alternative splice sites was exposed by a change of the oncogenic properties of Akv MLV, when synonymous mutations destroying the SA' site, the SD' site, or both sites simultaneously were introduced. First and foremost, the obvious capability of Akv MLV to induce follicular B cell lymphoma was seriously weakened, when one or both of the alternative splice sites were mutated, suggesting that this competence relies on intact SA' and SD' sites and a proper balance between all the produced transcripts (one full-length, two singlespliced, and one double-spliced). The integrity of Akv MLV seems fundamental for its capability to induce FBL; thus we just reported that Akv MLVs with mutated enhancer sequences retained the ability to induce tumors of B-cell type, but altered specificities were observed, including an impaired ability of FBL induction [11] A more complex picture was observed regarding the strong predisposition of Akv MLV for plasmacytoma induction. This predisposition was affected significantly only if the SA' alone was mutated. Thus, if the SD' site was mutated along with SA', wild-type potential was restored. This may indicate that the ability to induce plasmacytoma is dependent on a fine-tuned balance between the alternative single-spliced and double-spliced transcripts. If no double-spliced transcript is produced, while the singlespliced 4.4 kb transcript still is, as is the case for the SA' mutant, the single-spliced transcript somehow seems to be related with a barrier for plasmacytoma induction. On the other hand, if both transcripts are produced (Akv-wt) or none of them are produced (the SA'/SD' double mutant) the virus will hold a potential for inducing plasmacytomas. This is in line with the overall observation that the most pronounced effects were observed when the SA' or SD' splice sites were mutated individually, while the outcome of infection with the SA'/SD' double mutant in essence, except for the capability of FBL induction, was comparable to that of the wild-type virus. It may thus be speculated if a delicate balance between the alternatively single-spliced and double-spliced transcripts is a key determinant for the shift in oncogenic specificity, as demonstrated by the SA' and SD' splice site Akv mutants.
The most striking shifts in specificity observed were the increased tendency to stimulate development of splenic marginal zone lymphoma and the exposure of a novel ability for inducing histiocytic sarcoma for the SA' site mutant. Although we have shown that mutation of the SA' site results in inhibition of generation of the doublespliced product, we are at this point not able to explain or point to any detailed mechanisms underlying the observed changes in specificity. The other remarkable shift in specificity was detected with the SD' mutant, which was the only virus capable of inducing centroblastic DLBCL, i.e. tumors for which an origin from the follicle or marginal zone could not be inferred and comparable to de novo DLBCL in humans. Moreover, since the exposed potential appeared quite strong (35% of the SD'-mutant induced tumors fell within this diagnosis), and since such tumors are in general rare in mice, this mutant virus may be a helpful starting point tool to create a solid mouse model of human de novo DLBCL.
Obviously, the proposed significance of proper balances between the four different transcripts for the observed shifts in tumor specificity may reflect a need for a well-regulated balance between resulting translational products. We did not investigate if novel proteins were produced, but the open reading frames (ORFs) of both the alternatively singly and doubly spliced transcripts clearly reveal a potential for additional proteins to be produced. The gene products of the single-spliced 4.4 kb transcript most probably correspond to the p50 and p60 proteins made from the equivalent Friend MLV transcript [19]. These proteins were produced with translation initiations at two initiation codons (AUG gag and CUG glyco-gag ) in the same ORF and were shown to harbor the N-terminal Gag domain including matrix, p12, and the first 110 amino acids of the capsid in frame with the last 116 amino acids of integrase [19]. Also the smaller double-spliced transcript harbors smaller ORFs providing a scene for even more MLV proteins.
Intragenic elements such as gag enhancers have been known for many years in avian retroviruses [30,31]. However, it seems unlikely that a similar element is involved here, since the mutant virus with SA' and SD' sites mutated together was clearly the less affected one. This observation, in concert with the observed consequence on generation of different splice products, more likely suggests that the effect on disease specificity is related to an RNA processing phenomenon rather than an intragenic gag determinant with an effect on transcription.
Lymphoma-induction by non-acute murine retroviruses is associated with multiple proviral insertions that affect critical host genes. To achieve such multiple insertions the superinfection resistance caused by Env-expression must be by-passed. One possibility could be that reduced Envexpression caused by mutation of the gag splice sites as reported here might favor superinfection and thereby multiple proviral insertions. While we cannot exclude this possibility, our finding of the same number of sequence tags for proviral insertions for the wild-type and mutated viruses gives no immediately support to such a mechanism.
Retroviral insertional mutagenesis has been established as a solid strategy for the identification of candidate cancercausing genes [6,26,[32][33][34]. Accordingly, in an effort to relate specific genes or pathways with specific diagnoses, splice pattern, or causative virus, we identified a pool of 240 integration sites from which 111 novel RISs and 35 novel CISs were defined. Our analyses did not immediately point to any clear correlations; nevertheless the collection of candidate genes may prove to be a central input in future attempts to understand the exact roles of the different splice transcripts and/or their resulting translational products in hematopoietic differentiation and tumorigenesis.

Conclusion
We have in the B-lymphomagenic Akv MLV in the gag region identified a novel exon, which represents the first example of a doubly spliced gammaretroviral transcript.
Mutations of the alternative splice sites that define this novel transcript change the distribution of the different induced tumor phenotypes as well as generate tumors of additional specificities such as de novo diffuse large B cell lymphoma and histiocytic sarcoma. Provirus integration site analyses revealing 111 novel RISs and 35 novel CISs did not clearly point to specific target genes or pathways to be associated with specific tumor diagnoses or individual viral mutants. However, the list of potential target genes will be useful for future studies of hematopoietic differentiation and tumorigenesis.
Danish Natural Sciences and Medical Research Councils, NIH grant CA100266, Synergenics LLC, and the National Danish Research Foundation through the Centre for mRNP Biogenesis and Metabolism at the University of Aarhus.