Irf3 Polymorphism Alters Induction of Interferon Beta in Response to Listeria monocytogenes Infection

Genetic makeup of the host plays a significant role in the course and outcome of infection. Inbred strains of mice display a wide range of sensitivities to Listeria monocytogenes infection and thus serve as a good model for analysis of the effect of genetic polymorphism. The outcome of L. monocytogenes infection in mice is influenced by the ability of this bacterium to induce expression of interferon beta mRNA, encoded in mouse by the Ifnb1 (interferon beta 1, fibroblast) gene. Mouse strains that lack components of the IFNβ signaling pathway are substantially more resistant to infection. We found that macrophages from the ByJ substrain of the common C57BL/6 inbred strain of mice are impaired in their ability to induce Ifnb1 expression in response to bacterial and viral infections. We mapped the locus that controls differential expression of Ifnb1 to a region on Chromosome 7 that includes interferon regulatory factor 3 (Irf3), which encodes a transcription factor responsible for early induction of Ifnb1 expression. In C57BL/6ByJ mice, Irf3 mRNA was inefficiently spliced, with a significant proportion of the transcripts retaining intron 5. Analysis of the Irf3 locus identified a single base-pair polymorphism and revealed that intron 5 of Irf3 is spliced by the atypical U12-type spliceosome. We found that the polymorphism disrupts a U12-type branchpoint and has a profound effect on the efficiency of splicing of Irf3. We demonstrate that a naturally occurring change in the splicing control element has a dramatic effect on the resistance to L. monocytogenes infection. Thus, the C57BL/6ByJ mouse strain serves as an example of how a mammalian host can counter bacterial virulence strategies by introducing subtle alteration of noncoding sequences.


Irf3 Polymorphism Alters Induction of Interferon Beta in Response to Listeria monocytogenes Infection
Introduction Bacterial pathogens utilize a wide range of approaches to down-modulate or subvert host immune responses. L. monocytogenes is an intracellular pathogen that, following invasion of the host cell, is capable of escaping the host phagolysosomes and replicating in the cytoplasm. Within the cytoplasm, the bacterial DNA is thought to be recognized by an unknown host receptor, activating a signaling cascade that rapidly induces Ifnb1 expression [1]. This signaling cascade relies on TANK-binding kinase 1 (TBK1)-mediated phosphorylation of IRF3, a transcription factor that, following dimerization and translocation to the nucleus, induces expression of Ifnb1 [2][3][4].
In a murine model of infection, activation of host IFNb signaling is an important L. monocytogenes virulence strategy. Mouse lines that lack components of the IFNb signaling pathway (Ifnb1, Ifnar1) are significantly more resistant to L. monocytogenes infection [2,[4][5][6]. A similar protective effect of a Irf3 knockout suggests that Toll-like receptor (TLR)-independent induction of IFNb is detrimental to control of listeriosis [7]. Several independent observations suggested that IFNb signaling sensitizes lymphocytes for cell death, leading to an increase in sensitivity to L. monocytogenes [8]. L. monocytogenes activates such proapoptotic genes as Trail (Tnfsf10), Pkr (Eif2ak2), and Daxx in spleen and bone marrow macrophages of wild-type, but not Ifnar-deficient mice [4]. This is consistent with the observation that Trail knockout mice are more resistant to L. monocytogenes infection [9]. It has also been noted that following infection, mice lacking infection was carried out using a pair of differentially susceptible inbred mouse strains: BALB/cByJ and C57BL/ 6ByJ [11]. These strains were selected based on the ancestry of the 13-member CXB Recombinant Inbred (RI) Panel, which serves as a useful tool for mapping single gene traits [12]. While our study identified two major genetic loci that controlled differential sensitivity to L. monocytogenes infection, it was clear that there were additional genetic factors that we did not detect due to the limited size of our cross. Here, we report identification and characterization of one of these additional factors, a polymorphism in the C57BL/6ByJ inbred mouse strain that affects expression of Ifnb1 and results in increased resistance to L. monocytogenes infection. Our data demonstrate that a single base-pair polymorphism in intron 5 of Irf3 reveals an important role for splicing in control of IFNb induction and innate immune function.

C57BL/6ByJ Strain-Specific Defect in Induction of Ifnb1 Expression
Our analysis of L. monocytogenes infection of macrophages derived from bone marrow (i.e., bone marrow macrophages; BMMs) of BALB/cByJ and C57BL/6ByJ strains revealed strainspecific differences in infection-induced cell death. Across a range of time points and infectious doses, BALB/cByJ BMMs had consistently higher cell death then BMMs from C57BL/ 6ByJ mice ( Figure 1A and 1B). Interestingly, 18 h following infection with L. monocytogenes, there were also significant (p , 0.001) differences in death of BMMs from J and ByJ substrains of the common C57BL/6 lineage ( Figure 1C). The observed differences in cell death could be due to small differences in replication of bacteria in infected BMMs ( Figure 1D). However, recent studies that have demonstrated a role for IFNb signaling in the outcome of L. monocytogenes infection have also suggested that it plays a role in the survival of macrophages [2]. Therefore, we chose to test if there are mouse strain-specific differences in IFNb signaling, and we analyzed the course of Ifnb1 mRNA induction in L.
monocytogenes-infected BMMs. We found that Ifnb1 expression was rapidly induced in BALB/cByJ BMMs, increasing 400-fold by the 4-h time point. By contrast, in C57BL/6ByJ BMMs, there was very slight induction of Ifnb1 mRNA, reaching only 12-fold by the 4-h time point (Figure 2A). This low level of Ifnb1 induction in C57BL/6ByJ mice was surprising, as the role of Ifnb1 signaling in conferring susceptibility to L. monocytogenes infection has been analyzed using mice of the C57BL/ 6 background [4]. To test the possibility that the defect in Ifnb1 induction was specific to the C57BL/6ByJ substrain we also monitored expression of Ifnb1 in BMMs from C57BL/6J mice. Indeed, in response to L. monocytogenes infection, C57BL/ 6J BMMs had a similar magnitude of Ifnb1 mRNA induction as BALB/cByJ BMMs (Figure 2A). Ifnb1 mRNA induction in F1 progeny of the J and ByJ substrains of mice was similar to that observed in the C57BL/6J strains (unpublished data), suggesting that the C57BL/6ByJ strain carries a recessive polymorphism that prevents upregulation of Ifnb1 expression. There is at least one additional inbred mouse strain, SPRET/ Ei, that has a similar naturally occurring defect in induction of Ifnb1 expression [13]. However, the genetic basis for this defect in SPRET/Ei remains to be elucidated.
Differential Sensitivity of C57BL/6 Substrains to L. monocytogenes Infection Several lines of evidence indicate that inhibition of IFNb signaling in mice leads to a significant increase in resistance to intravenous L. monocytogenes infection [4,5]. To test if the Ifnb1 induction defect in the C57BL/6ByJ strain promotes resistance to infection, we compared survival and bacterial loads in C57BL/6 substrains infected intravenously with a high dose (10 5 colony-forming units [cfu]) of L. monocytogenes strain 10403S. There was no detectable Ifnb1 expressed in liver and spleen tissue in either mouse strain up until the 24-h time point. At the 24-h time point, Ifnb1 mRNA was present in both livers and spleens of C57BL/6J animals but only in spleens of C57BL/6ByJ animals ( Figure S1). We found that at the 24-and 48-h time points, there was a significantly higher number of bacteria in the livers ( Figure 3A) and spleens ( Figure 3B) of C57BL/6J mice, indicating the infection was controlled better in C57BL/6ByJ mice. Consistent with this idea, spleens from C57BL/6ByJ animals had nearly twice as many Mac1-positive cells than spleens from C57BL/6J mice at the 48-h time point ( Figure 3C). In fact, this dose of bacteria led to death in C57BL/6J animals within 72 h after infection, whereas most of the C57BL6/ByJ mice survived indefinitely ( Figure S2). Collectively, these results indicate that C57BL/ 6ByJ mice were significantly more resistant to intravenous L. monocytogenes infection than C57BL/6J animals.

Ifnb1 Induction Defect Lies in a Shared Component of the Signaling Pathway
Intracellular L. monocytogenes is thought to induce Ifnb1 expression by activating an as-of-yet unidentified cytoplasmic receptor that initiates signaling through the TBK1 and inhibitor of kappaB kinase epsilon (IKBKE) kinases [1]. TBK1 and IKBKE also participate in transducing signals from various TLRs in response to viral and bacterial infections [14]. Therefore, we tested if the defect in C57BL/ 6ByJ mice was in a L. monocytogenes-specific component of the

Author Summary
Specific variances in an individual's DNA, known as genetic polymorphisms, can play a significant role in determining susceptibility to an infectious disease. To identify the genetic polymorphisms that are associated with resistance to the common human bacterial pathogen L. monocytogenes, we have carried out a series of genetic and molecular biology experiments using closely related strains of mice that are differentially susceptible to Listeria infection. Through this analysis, we have identified a spontaneous mutation in an intron of the Irf3 gene, which encodes a key transcription factor involved in innate immunity. This single nucleotide change affects the efficiency with which Irf3 mRNA is spliced, thus limiting the ability of bacteria to induce interferon beta expression in order to suppress innate immune defense. By analyzing this mutation, we found that processing of mouse Irf3 mRNA relies on an atypical U12 splicing mechanism that has been suggested to be a rate-limiting step in gene expression. Our findings not only provide an additional example of an important role of noncoding polymorphisms in control of gene function, but also demonstrate how such polymorphisms can fine tune innate immune response.
lipopolysaccharide or poly I:poly C, which induce Ifnb1 expression through TLR4 and TLR3, respectively, failed to induce Ifnb1 mRNA at the same levels as observed in C57BL/6J BMMs ( Figure 2B and 2C). On the other hand, C57BL/6ByJ BMMs treated with 200 hemagglutinizing units (HU) of Sendai virus, which induces Ifnb1 expression through a RIG-I/MAVS-dependent pathway [15], had levels of Ifnb1 mRNA comparable to those observed in C57BL/6J BMMs at the later stages of infection but nevertheless had a noticeable delay at the earlier stages ( Figure 2D). These observations indicate that the defect in Ifnb1 induction in C57BL/6ByJ mice is likely to lie in a shared component of the signaling pathway. However, our initial analysis Tbk1, Ikbke, and Irf3 failed to identify differences in the coding sequence or overall expression levels of these mRNAs in C57BL/6ByJ versus C57BL/6J mice (unpublished data).

Mapping of the Ifnb1 Induction Trait
As mentioned above, our original choice of mouse strain for genetic analysis was based on the availability of the CXB RI mapping panel. However, analysis of the transcriptional response to L. monocytogenes infection in macrophages from all 13 CXB strains revealed no differences in induction of Ifnb1 (unpublished data), precluding the use of the panel for mapping. Because C57BL/6ByJ mice carry a recessive muta-tion, we therefore chose a backcross as our mapping strategy to identify the locus in the C57BL/6ByJ mouse genome that harbors the mutation preventing induction of Ifnb1. C57BL/ ByJ and C57BL/6J mice have virtually no polymorphisms that can be used to monitor allelic segregation in a cross. On the other hand, BALB/cByJ mice are similar to C57BL/6J mice in their induction of Ifnb1, and we therefore chose C57BL/6ByJ and BALB/cByJ as parental strains for our cross. We backcrossed F1 male progeny of C57BL/6ByJ and BALB/cByJ mice to C57BL/6ByJ females and used the resulting 54 C(B.C) N2 progeny to construct a genetic map with 56 microsatellite markers evenly distributed throughout the mouse genome [16]. To generate phenotypic data, we first used real time reverse transcriptase PCR (RT-PCR) to analyze the dynamics of L. monocytogenes-induced Ifnb1 expression in BMMs isolated from 43 backcrossed mice (unpublished data). We then used transformed real time RT-PCR Ct values representing the levels of Ifnb1 mRNA at the 4-h time point directly as a quantitative trait. Mapping of this trait using MapManger QTX identified a peak likelihood ratio statistic score of 35.1 (logarithm of the odds [LOD] ¼ 7.6) at the D7Mit229 marker [17] ( Figure S3). Using the MapManger QTX built-in permutation test function, we established that the identified linkage is highly significant with an experimental p-value less then 10 À4 . QTL support interval was approximated as 1.5 LOD drop-off from the peak score [18] and extended from D7Mit27 to D7Mit158 markers, spanning empirical genetic distance of 6.9 cM (0 cM MGI) and a physical fragment of 2.35 Mb. The D7Mit229 marker is located on mouse Chromosome 7 adjacent to Irf3 [16], identifying Irf3 as the primary candidate gene.

Irf3 mRNA in C57BL/6ByJ Strain Is Not Efficiently Spliced
Interestingly, our initial analysis of C57BL/6J and C57BL/ 6ByJ substrains did not find any differences in overall Irf3 mRNA levels or polymorphisms in the coding region of Irf3 (unpublished data). However, analysis of the structure of Irf3 mRNA using a series of overlapping primers revealed differences in the Irf3 transcripts between the two substrains. As expected, in C57BL/6J mice the majority of Irf3 transcripts were completely spliced, whereas in C57BL/6ByJ mice, splicing of Irf3 was not complete and the majority of transcripts retained intron 5 ( Figure 4A and quantified in Figure 4B). Retention of intron 5 introduces a premature stop codon at amino acid 243, rather than producing the fulllength 419 amino-acid protein. To test if the observed differences in splicing had functional significance, we analyzed activation of IRF3 in BMMs by monitoring the formation of IRF3 dimers following bacterial (L. monocytogenes) or viral (Sendai virus) infection. We found that untreated BMMs from C57BL/6ByJ mice had significantly lower levels of IRF3 protein than untreated BMMs from C57BL/6J mice ( Figure 4C and 4D). Moreover, following a 2-h infection with L. monocytogenes or Sendai virus, there were no detectable IRF3 dimers in C57BL/6ByJ BMMs, although IRF3 dimers were readily detectable in C57BL/6J BMMs. Nevertheless, we observed that Sendai virus-infected C57BL/6ByJ BMMs are capable of inducing Ifnb1 expression (see Figure 2D). This is consistent with earlier observations that Irf3-deficient cells rely on IRF7 to have a normal interferon response to several viral infections [19,20]. Interestingly, our polyclonal antibodies failed to detect a truncated form of IRF3 even in the presence of proteasome inhibitor (MG132), suggesting that the unspliced form of Irf3 might not be efficiently translated ( Figure 4C). Overall, these results show that C57BL/6ByJ BMMs have dramatically lower amounts of functional IRF3 protein, and in conjunction with the existing Irf3 knock-out data [2,4], explain the increased resistance of C57BL/6ByJ mice to L. monocytogenes infection (see Figure 3). A to T Polymorphism Impairs Slicing of Irf3 Intron 5 in C57BL/6ByJ Strain Sequencing of the entire 7.2-kb genomic region of Irf3 [21], including 1 kb of upstream and downstream sequences, revealed a single A to T polymorphism in the middle of intron 5 in C57BL/6ByJ mice ( Figure 5). To establish if this polymorphism altered the splicing efficiency of the intron, we monitored splicing using both cell culture-based and in vitro approaches. For cell culture-based experiments, we derived minigene constructs containing the complete intron 5 (from either C57BL/6J or C57Bl/6ByJ) flanked by exons 5 and 6, and expressed them under the control of the heterologous CMV promoter. In order to rule out the possibility that C57BL/6ByJ mice carry additional mutations that affect splicing, we first tested our constructs in a C57BL/6ByJ fibroblast-like cell line (Y5). Following transfection into Y5 cells, the efficiency of splicing of intron 5 was monitored by real time RT-PCR using primers specific to the vector and exon 5-6 junctions, and the total amount of RNA expressed from each construct was measured using primers specific to the exon fragment, which is identical in both constructs (see Figure 6A schematic). When normalized for the total amount of expressed RNA, there was significantly more spliced product generated from the C57BL/6J construct than from the C57BL/6ByJ construct ( Figure 6A). This result indicates that in C57BL/6ByJ cells, C57BL/6J Irf3 intron 5 is spliced more efficiently than the C57BL/6ByJ version of the intron.
The effect of the A to T substitution on splicing efficiency of intron 5 was further confirmed using an in vitro splicing assay, in which a uniformly radioactively labeled Irf3 pre-mRNA containing intron 5 flanked by 50 bp of exon 5 and exon 6 was incubated in HeLa nuclear extract. As expected, the C57BL/6J-derived Irf3 pre-mRNA substrate was spliced efficiently, as evidenced by the appearance of both intermediate and fully spliced products ( Figure 6B). By contrast, there was no detectable splicing of the C57BL/6ByJ-derived Irf3 pre-mRNA substrate even when incubated for 60 min in the splicing reaction mixture. These results confirmed that the A to T substitution had a direct effect on efficiency of Irf3 splicing.

Splicing of Murine Irf3 Intron 5 Relies on U12 Splicesome
Pre-mRNA splicing occurs in a ribonucleoprotein complex called the spliceosome [22]. Splicing is initiated through recognition of several intron-defining splicing signals, including the 59 and 39 splice sites and the branchpoint, which is usually located near the 39 end of the intron. Introns can be classified into two categories: U2-type introns, which comprise the major class of introns, and U12-type introns. U2type introns are characterized by the presence of a conserved GU dinucleotide at the 59 end of the intron and a conserved AG dinucleotide at the 39 end, whereas U12-type introns can harbor AT-AC, AT-AG, and GU-AG dinucleotides at their 59 and 39 boundaries, respectively. Furthermore, on U12-type introns, the 59 and 39 splice sites and branchpoint are highly conserved and differ from those of the conventional U2-type introns [23,24], and the characteristic polypyrimidine tract is typically absent [25].
Because the A to T polymorphism is located within an intron and affects splicing efficiency, we hypothesized that it might alter the function of a splicing signal. The poly-morphism is located 46 bp upstream of the 39 splice site, within a region where the branchpoint is typically found. The 59 boundary of the murine Irf3 intron 5 matches the U2-type intron GTRAGT consensus sequence ( Figure 5) [23,24]. However, the region surrounding the polymorphism more closely resembled the U12 branchpoint consensus (TCCTTA-ACY) than the U2 branchpoint consensus (YURAY). To determine whether intron 5 of the Irf3 gene was a U2-type or U12-type intron, we monitored splicing of the wild-type C57BL/6J-derived Irf3 pre-mRNA substrate following inactivation of U2 or U12 snRNA by oligonucleotide-directed RNase H cleavage. The in vitro splicing assay in Figure 6C shows that splicing of the C57BL/6J-derived Irf3 pre-mRNA substrate occurred following inactivation of U2 snRNA but not following inactivation of U12 snRNA. As expected, splicing of the control U2-type intron-containing adenovirus major late (Ad ML) pre-mRNA substrate was fully dependent on the presence of U2 snRNA. These observations indicate that splicing of intron 5 of the Irf3 gene relies on the U12dependent mechanism.

Irf3 Intron 5 A to T Polymorphism Affects Induction of Ifnb1 Expression
Irf3 transcripts that retain intron 5 are detected in both C57BL/6ByJ and C57BL/6J strains (see Figure 4A). Therefore, it appears that even in the C57BL/6J strain splicing of intron 5 is somewhat inefficient, whereas in the C57BL/6ByJ strain, intron 5 splicing is substantially impaired. To test if the observed phenotypic differences in induction of Ifnb1 expression in the two mouse strains are due to differences in splicing efficiency of Irf3 intron 5, we performed complementation experiments. To achieve this, we transfected BMMs from Irf3 knockout mice with full-length in vitro transcribed Irf3 mRNA species harboring either the C57BL/6J or C57BL/6ByJ version of intron 5. Fully spliced Irf3 mRNA was used as a positive control, and mRNA containing a partial deletion of IRF domain (DXma) was used as a negative control. Previous experiments had shown that introduction of single-stranded RNA into the cell cytosol leads to Irf3dependent induction of Ifnb1 expression [26][27][28] (O. G., unpublished data). Therefore, transfection of intron 5containing Irf3 mRNAs into BMMs that lack Irf3 should lead to a level of Ifnb1 induction that is proportional to the splicing efficiency of intron 5. We measured the levels of Ifnb1 mRNA by real time RT-PCR 19 h following transfection of BMMs, and found that there was ;5-fold more (p , 0.01) Ifnb1 mRNA expressed in BMMs transfected with the C57BL/6Jderived Irf3 mRNA than in BMMs transfected with C57BL/ 6ByJ-derived Irf3 mRNA (Figure 7). Following a 4-h infection with L. monocytogenes, BMMs transfected with either mRNA showed further induction of Ifnb1 expression. Nevertheless, the C57BL/6J-derived Irf3 mRNA induced significantly (p , 0.01) higher amounts of Ifnb1 mRNA than the C57BL/6ByJderived version. As expected, BMMs transfected with Irf3 mRNA lacking the IRF domain induced only low levels of Ifnb1 mRNA expression. Because the Irf3 knockout mice used in this experiment are on the C57BL/6J background [19], this experiment also ruled out the possibility that impaired splicing of Irf3 intron 5 could be due to a linked C57BL/ 6ByJ polymorphism. Therefore, we conclude that decreased splicing efficiency of Irf3 intron 5 is directly responsible for the reduction in Ifnb1 expression observed in C57BL/6ByJ mice.

Discussion
Previous studies have demonstrated an important role of IFNb signaling in host defense against L. monocytogenes infection. L. monocytogenes evolved to take advantage of the host signaling pathways and is capable of inducing Ifnb1 expression in order to down-modulate the antibacterial host defense. Here, we show that at least one inbred strain of mice can resist this pathogen's tactic by carrying a single nucleotide polymorphism that changes the efficiency of splicing of its Irf3 transcription factor. While this naturally occurring polymorphism does not eliminate IRF3 activity, the resulting reduction in IRF3 protein levels is sufficient to confer 10-fold higher resistance to L. monocytogenes infection. Considering that complete loss of IRF3 function is detrimental to immune defense, and Irf3 knockout mice are more sensitive to encephalomyocarditis infection [19], it would be interesting to determine if the level of IRF3 in C57BL/6ByJ mice is sufficient to maintain protection against viral infections. Nevertheless, our finding indicates that genetic changes in noncoding regions of the host genome is one of the mechanisms that can be used to fine tune the effectiveness of host defenses against infections.
It has been suggested that in the process of evolution, U12type introns are either lost or undergo subtype switching (from AT-AC to GT-AG) and are eventually converted to U2type introns [25]. Our data provide additional support for this hypothesis. Although splicing of IRF3 intron 5 is dependent on the U12 spliceosome, the splice donor site is a typical U2 site. Even more interesting is the fact that the region of the human and rhesus intron 5 that is homologous to the putative murine U12 branchpoint site contains a G in place of the T that is found in rodents ( Figure 5, inset box). This substitution places a purine residue in front of a putative branchpoint adenosine, thus creating a perfect match to the canonical U2 branchpoint consensus sequence. Therefore, intron 5 of the murine Irf3 gene might represent one of the final steps in the conversion of a U12-type intron to a U2-type intron.
The amount of IRF3 available in the cell is tightly controlled and overproduction of IRF3 is lethal to BMMs Figure 5. Alignment of the Intron 5 Sequences from the Irf3 Genes from Mouse (Substrains C57BL/6J and C57BL/6ByJ), Rat, Human, and Rhesus Monkey The A to T polymorphism (arrow) observed in the Irf3 gene in C57BL/6ByJ mice is found within a conserved sequence (boxed) similar to the U12 TCCTAACY branchpoint consensus. A T to G substitution found in the human and rhesus genes converts a pyrimidine to a purine residue in front of a putative branchpoint adenosine and creates a perfect match (inset box) to the U2 YURAY branchpoint consensus. doi:10.1371/journal.pgen.0030152.g005 Figure 6. Mouse Strain-Specific Differences in Irf3 Intron 5 Splicing (A) Relative levels of the spliced to unspliced forms of an Irf3 intron 5-containing minigene expressed in C57BL/6ByJ fibroblast-like Y5 cells 12 h following transfection. Levels are normalized to the total amount of expressed RNA. The location of the primers used in the analysis is shown in the schematic diagram. The efficiency of splicing of intron 5 was monitored using primers specific to the vector and exon 5-6 junctions (top pair), whereas the total amount of RNA expressed from each construct was measured using primers specific to the exon fragment (bottom pair). (B) In vitro splicing of the C57BL/6J of C57BL/6ByJ minigene pre-mRNA was analyzed at 0, 15, 30, and 60 min following addition of the substrate to the nuclear extract. Identities of the spliced products are shown of the left. (C) In vitro splicing of the C57BL/6J minigene pre-mRNA was analyzed at 0, 30, and 60 min following inactivation of U2 or U12 snRNA by oligonucleotide directed RNase H cleavage. As a control, splicing of the U2-type intron-containing adenovirus major late (Ad ML) pre-mRNA substrate was analyzed. doi:10.1371/journal.pgen.0030152.g006 [29,30] (Figure S4). Recent evidence indicates that splicing of U12-type introns could be a rate-limiting step in gene expression [31]. Our analysis suggests that this may also be the case for intron 5 in the murine Irf3 gene. Macrophages from common strains, such as C57BL/6J and BALB/cByJ, have two populations of Irf3 transcripts: a major, fully spliced species and a minor species that retains intron 5 (see Figure  4A). The presence of multiple Genbank (i.e., BC082274, BC003233; http://www.ncbi.nlm.nih.gov/Genbank/index.html) and EST entries of Irf3 mRNAs that retain intron 5 further supports this hypothesis. Therefore, the rate of intron 5 splicing could control the amount of IRF3 available in the cell. Interestingly, it has been shown that activity of human IRF3 is also regulated at the level of splicing [32]. However, in contrast to rodent Irf3, regulation of human IRF3 involves alternative intron 1 splice acceptor sites that can produce an active or a dominant negative, IRF3a, form of IRF3 [33]. Because it appears that by converting to a U2 intron, human IRF3 intron 5 lost its rate-limiting function, it is intriguing to contemplate that this led to emergence of an alternative splicing control mechanism for human IRF3.
Our choice of mouse strains for genetic analysis of susceptibility to L. monocytogenes infection was based on the existence of the ByJ-based CXB RI mapping panel. The C57BL/6By substrain has been used to generate at least seven of the 13 CXB RI strains, but none of the CXB strains appear to have a defect in induction of Ifnb1. To resolve this discrepancy, we sequenced Irf3 intron 5 in all C57BL/6Byderived CXB strains. None of the sequenced Irf3 introns contained the mutation found in the C57BL/6ByJ strain (unpublished data). Therefore, it appears that the A to T mutation rose in the C57BL/6ByJ background only recently, after the generation of the CXB RI strains. It is possible that the return of D. W. Bailey's substrains to the Production Department of Jackson Laboratories in 1974 could have created a bottleneck that fixed the mutation in the current C57BL/6ByJ population.
Splicing of mRNA is a critical step in protein expression, and in humans, genetic polymorphisms that produce aberrant or alternate splicing products have been associated with a wide range of diseases [34]. We used genetic analysis of the mouse model system to provide definitive evidence of the important role of splicing in control of infection. We found that a mouse substrain-specific defect in induction of Ifnb1 is due to a single nucleotide polymorphism in intron 5 of Irf3. Our analysis of this polymorphism revealed that splicing is a critical step in the control of Irf3 expression and, as a result, in the course and outcome of L. monocytogenes infection. While intron 5 of murine Irf3 has features of both U2 and U12 introns, we provide evidence that its splicing is dependent on the U12 spliceosome. Therefore, it appears that in rodents the U12 spliceosome can use U2 splice sites. This suggests that the spectrum of U12-type introns present in mammalian genomes could be wider than previously thought. Finally, our comparison of rodent and primate Irf3 genomic sequences also revealed the intriguing possibility that we have identified an intermediate step in the process of conversion from a U12-to U2-type intron.

Materials and Methods
Animals. Six-to-twelve-week-old animals were used in all experiments. BALB/cByJ, C57BL/6J, and C57BL/6ByJ mice were obtained from Jackson Laboratories (http://www.jax.org/). B/ByJ.C N2 mice were created by backcrossing B/ByJ.C F1 males to C57BL/6ByJ females. All mouse strains were bred and maintained under specific pathogenfree conditions in the animal facilities at the University of Massachussetts Medical School. All experiments involving live animals were carried out in accordance with the guidelines set forth by the University of Massachussetts Medical School Department of Animal Medicine and the Institute Animal Care and Use Committee.
In vivo infections. Pre-titered TSB-glycerol stocks of L. moncytogenes strain 10403S were stored at À80 8C. Prior to infection, 1-ml bacterial aliquots were recovered for 1 h at 37 8C in 9 ml of TSB (BD Biosciences, ttp://www.bdbiosciences.com/), washed, and resuspended to the desired cfu in PBS. Mice were injected with a defined dose of L. monocytogenes strain 10403S in 0.4 ml of PBS. At defined time points, infected animals were killed by CO 2 asphyxiation. Livers and spleens of infected animals were aseptically harvested, weighed, and homogenized in 0.02% Triton X-100. Aliquots of serial 5-fold dilutions in sterile water were plated in duplicate on TSB agar (BD Biosciences) plates containing 10 lg/ml streptomycin. After overnight incubation, the number of bacteria per milligram of tissue was determined by counting colonies at the appropriate dilution.
Cell lines. Fibroblast-like YF5 and macrophage-like YM14 cell lines were generated by immortalization of C57BL/6ByJ bone marrow cells as previously described [35,36]. Clones were selected based on morphology.
Ex vivo experiments. Differentiated BMMs were detached from the Petri plates by incubation in cold PBS, washed, resuspended in BM medium without antibiotics and used to seed multiwell dishes at 1 3 10 5 cells/cm 2 . Following overnight incubation the medium was replaced with DMEM containing the agent used to stimulate BMM. For L. monocytogenes, after 1 h the medium was replaced with BM medium containing 10 lg/ml gentamicin (Fisher) to remove extracellular bacteria. Lipopolysaccharide (Sigma) was added at 1 lg/ml, poly I:poly C (Sigma) at 25 lg/ml and Sendai Virus (generous gift of Kate Fitzgerald, University of Massachussetts) at 200 U/ml. At defined timepoins BMMs were lysed in TRIzol (Invitrogen) and RNA was isolated according to manufacturer's protocol.
BMM cell death assay. 5 3 10 4 BMMs were seeded in 96-well tissue culture plate and following overnight incubation the cells were infected with a defined multiplicity of infection (MOI) of L. monocytogenes. After 1 h incubation, the cell culture medium was replaced with medium containing 10 lg/ml of gentamicin (MP Biomedicals, http://www.mpbio.com/). At defined time points the supernatants were collected and the remaining cells were lysed for 20 min at room temperature in 50 ll/well of complete medium plus 1ll of cell lysis solution from CytoTox-ONE Homogeneous Membrane Integrity Assay kit (Promega, http://www.promega.com/). Lactose dehydrogenase (LDH) assay for supernatants and cell lysates was performed according to the manufacturer's protocol. Fluorescence was recorded at 560/590 nm using a Synergy HT microplate reader (Bio-Tek, http://www.biotek.com/). The relative concentration of LDH in supernatants was calculated by the equation: 100% 3 (LDHsup/ (LDHsupþLDHcell)).
Real time PCR. Relative mRNA levels were quantified by real time RT-PCR on an ABI 7300 instrument (http://www.appliedbiosystems. com/) utilizing SYBR Green chemistry (ABI SYBR master mix þ RT reagents). Generally, 50 ng of total RNA was used in 20 ml one-step reactions that incorporated a 30-min reverse transcription step prior to cycling. Primers used to detect specific mRNAs are described in Table S1. Ribosomal protein S17 (Rps17) and actin mRNA were used as a loading controls. To quantify the relative amounts of mRNA in each experiment, 2-fold dilutions were used to create calibration curves. Each experiment included at least two biological and three experimental replicates.
Genetic mapping. BMMs from 46 B/ByJ(C.B/ByJ) N2 mice were infected in duplicate with L. monocytogenes strain 10403S at MOI ¼ 5. Four hours following infection, total RNA was isolated and used for real time RT-PCR analysis of Ifnb1 mRNA induction. Ifnb1 Ct values from duplicate samples were adjusted for variation in total RNA concentration using Rps17 Ct values, transformed by subtracting the average C57BL/6ByJ parental value, and used directly as trait values for mapping using MapManger software. Under ideal conditions, such transformed Ct values can be viewed as log 2 of the fold difference in Ifnb1 expression compared to the C57BL/6ByJ parent. A genetic map was constructed using 56 microsatelite markers [16]. Experimental p-value for linkage was evaluated using a built-in permutation test and was found to be less than 10 À4 .
IRF3 immunoblotting. BMMs were infected with L. monocytogenes (MOI ¼ 5) or Sendai Virus (600 HU) for 4 h. Cells were lysed in the presence of protease inhibitors (Roche, http://www.roche.com/) and following centrifugation, supernatant aliquots containing 20 lg of protein were loaded per lane of a native polyacrylamide gel. IRF3 was detected by immunoblotting with an anti-IRF3 antibody (Invitrogen). For total IRF3 protein analysis, lysates were obtained from infected and uninfected BMMs treated with either 10 lM MG-132 proteasome inhibitor or DMSO vehicle. Three micrograms of total protein were separated on denaturing SDS polyacrylamide gel and immumoblotted with anti IRF-3 antibobody (Invitorgen).
Splicing assays. C57BL/6J and C57BL/6ByJ minigene constructs were generated by cloning Irf3 intron 5 flanked by 250 bp of exonic sequences into the vector pcDNA3 (Invitrogen), to create pINT5J and pINT5ByJ, respectively. For ex vivo splicing assays, the C57BL/6ByJderived fibroblast-like cell line Y5 was transfected with either pINT5J or pINT5ByJ, and splicing was monitored 12 h later by real time RT-PCR using primers specific to the transcribed vector sequence (pCDNA3F3) and either intron 5 (IRF3IntR) or the exon 5-6 junction (IRF3Ex5/6R5,). For in vitro splicing assays, minigene Irf3 pre-mRNA templates were PCR-amplified from pINT5J and pINT5ByJ using a T7-containing primer, purified, and transcribed in vitro using T7 polymerase in the presence of [alpha-32 P]UTP. In vitro splicing reactions were performed essentially as described previously [37], except that 30% HeLa nuclear extract was used. Spliced products were resolved on 12% denaturing polyacrylamide gels (19:1) in 8 M urea in Tris-Borate-EDTA buffer, and visualized using a Fujifilm FLA-500 phosphorimager (http://www.fujifilm.com/). U2 and U12 snRNAs were inactivated by RNase H-directed cleavage as described previously [38] using DNA oligonucleotides complementary to nucleotides 27-49 of the U2 snRNA or to nucleotides 11-28 of the U12 snRNA.