Regulation of p110δ PI 3-Kinase Gene Expression

Background Despite an intense interest in the biological functions of the phosphoinositide 3-kinase (PI3K) signalling enzymes, little is known about the regulation of PI3K gene expression. This also applies to the leukocyte-enriched p110δ catalytic subunit of PI3K, an enzyme that has attracted widespread interest because of its role in immunity and allergy. Principal Findings We show that p110δ expression is mainly regulated at the transcriptional level. In fibroblasts, lymphocytes and myeloid cells, p110δ gene transcription appears to be constitutive and not subject to acute stimulation. 5′RACE experiments revealed that p110δ mRNA transcripts contain distinct upstream untranslated exons (named exon -1, -2a, -2b, -2c and -2d), which are located up to 81 kb upstream of the translational start codon in exon 1. The levels of all the different p110δ transcripts are higher in leukocytes compared to non-leukocytes, with the p110δ transcript containing exon -2a most abundantly expressed. We have identified a highly conserved transcription factor (TF) binding cluster in the p110δ gene which has enhanced promoter activity in leukocytes compared to non-leukocytes. In human, this TF cluster is located immediately upstream of exon -2a whilst in mouse, it is located within exon -2a. Conclusion This study identifies a conserved PIK3CD promoter region that may account for the predominant leukocyte expression of p110δ.


Introduction
Phosphoinositide 3-kinases (PI3Ks) generate lipid second messengers that regulate a broad variety of cellular responses such as growth, cell cycle progression, differentiation, vesicular traffic and cell migration [1]. PI3K activity is critical in a wide variety of normal and pathological physiological responses, including immune regulation, metabolic control and cancer. However, despite the importance of this signalling system, very little is known about the regulation of PI3K gene expression under normal and disease conditions. The PI3K family is divided into 3 classes [2]. Class I PI3Ks are acutely activated upon receptor stimulation and are heterodimers consisting of a p110 catalytic subunit in complex with a regulatory subunit. The class I PI3Ks are further subdivided into class IA and IB, depending on whether the catalytic subunit is in complex with an SH2-domain containing regulatory subunit (collectively called 'p85') or with the p101 or p84 regulatory subunits, which lack SH2 domains. Mammals have 3 class IA p110 catalytic subunits, p110a, p110b and p110d, encoded by 3 distinct genes, PIK3CA, PIK3CB and PIK3CD, respectively. These p110 isoforms interact with p85, of which there are at least five different species, called p85a, p55a and p50a (encoded by the PIK3R1 gene) and p85b and p55c (encoded by PIK3R2 and PIK3R3, respectively). p110c is the only class IB PI3K catalytic subunit and occurs in complex with p101 [3,4] or p84 [5,6], which have no homology to p85. Class I PI3Ks can be activated by tyrosine kinases (p110a, p110d) or GPCRs (p110b and p110c) [1,[7][8][9].
Tissue distribution and the regulation of PI3K expression has recently been reviewed [10]. Whereas p110a and p110b appear to have a broad tissue distribution [11][12][13][14][15], p110d is highly expressed in leukocytes [12,13,16], found at intermediate levels in neurons [17] and present at low levels in most other cell types [13,18]. p110d is also expressed at moderate levels in some cancer cells of non-leukocyte origin such as melanoma and breast cancer cells, often with large differences in expression levels in cell lines of the same tissue origin [18], for reasons that are unclear at the moment. Like p110d, p110c is highly enriched in leukocytes [19][20][21] but is also found at lower levels in other cell types such as cardiomyocytes [22][23][24], endothelial cells [25], pancreatic islets [26,27] and smooth muscle cells [28].
Three recent studies have identified a transcription regulatory region for the human p110a gene, PIK3CA. The PIK3CA locus gives rise to two alternative transcripts which each contain a distinct 59 untranslated exon (exon -1b or -1a) that is spliced onto the first translated (ATG-containing) exon. The genomic position of these 59 untranslated exons is about 50 kb upstream of the translation start site [45,46]. TF binding sites for p53 [45]), FOXO3a [46]) and NF-kB [47] have been mapped in close proximity to the most 59 untranslated exon (called exon -1b). Whereas p53 might inhibit transcription of p110a, evidence for a positive regulation by NF-kB and FOXO3a has been presented.
A promoter region for murine p110c has also been identified [19]. Multiple transcriptional start sites exist for p110c, resulting in transcripts with varying 59 untranslated regions (59UTRs), up to 874 bp in length. Analysis of the genomic p110c DNA up to 1.2 kb upstream from the transcription start site revealed that the putative promoter region contains consensus sites for housekeeping TFs such as AP1 and SP1, as well as several putative binding sites for leukocyte-specific TFs [19]. Functional analysis of this p110c putative promoter region revealed enhanced promoter activity in the U937 myeloid cell line compared to the HeLa epithelial cell line [19].
In this study, we have investigated the regulation of p110d gene expression. We have documented that p110d protein expression largely correlates with the level of p110d mRNA in numerous cell types, indicating that p110d expression is predominantly regulated at the level of transcription. We have found multiple mouse and human p110d transcripts that contain distinct upstream untranslated exons, which we have named exon -1, -2a, -2b, -2c and -2d, located up to 81 kb upstream of the translational start codon in exon 1. Furthermore, we have identified a highly conserved TFbinding cluster that is located within mouse exon -2a and located immediately 59 upstream of human exon -2a. This TF-binding cluster has enhanced promoter activity in leukocytes compared to non-leukocytes. Out of the 7 different TF binding sites in the TFbinding cluster, 4 are associated with regulation of haematopoiesis and expression of leukocyte-specific genes. These findings are the first to identify a PIK3CD promoter and offer a rationale for the leukocyte-enriched expression of p110d.

p110d protein expression is not altered in fibroblasts, Blymphocytes and myelomonocytic cells upon acute stimulation with various agonists
We first investigated whether p110d expression can be induced by several acute cellular stimuli. In NIH-3T3 fibroblasts, which contain very low levels of endogenous p110d compared to leukocytes, p110d could not be induced by TNF, the proteasome inhibitor PS-341, UV irradiation, osmotic stress or the glucocorticoid dexamethasone (data not shown). p110d protein levels were also unaffected during different phases of the cell cycle in these cells (data not shown). In B lymphocytes, p110d expression was not affected by stimulation of the antigen receptor using anti-IgM antibodies. In U937 myelomonocytic cells, p110d levels were unaltered by treatment with retinoic acid, in contrast to the p110c protein which was induced effectively (data not shown), the latter in line with previously published data [48,49]. Taken together, p110d expression appears not to be regulated in an acute manner in response to extracellular stimuli, at least in the cell types and conditions investigated.

Correlation between p110d mRNA and protein levels in cell lines
We next assessed the levels of p110d protein and mRNA, using immunoblotting of total cell lysates and real time RT-PCR, respectively, in a panel of murine and human cell lines ( Figure 1A,B). p110d mRNA and protein were found in all cell lines investigated but in widely varying amounts. In line with published data [12,13,16,18], leukocytes expressed high levels of p110d while non-leukocytes expressed intermediate to low levels.
In line with previous data [12], a good correlation was found between p110d mRNA and protein levels in most cell lines tested, indicating that p110d protein expression is mainly regulated at the level of transcription.
DNA methylation and histone acetylation are unlikely to be key mechanisms to control PIK3CD expression DNA methylation and histone acetylation are important epigenetic mechanisms that control gene expression by dictating transitions between transcriptionally active or transcriptionally silent chromatin states [50,51]. L929 fibroblasts, which express low levels of p110d mRNA and protein, were treated with 59azacytidine or trichostatin A, agents known to cause DNA (hemi-) demethylation and histone hyperacetylation, respectively, creating open configurations of genomic DNA to allow binding of TFs. As a positive control, we monitored the previously documented induction in these cells of mRNA expression of the cytokines IL-6 and IFN-b by 59-azacytidine and trichostatin A [52,53]. As can be seen from Figure 2, p110d mRNA expression levels were not dramatically altered by any of these treatments, with a maximum increase in p110d mRNA of around 2-fold, which was not accompanied by an induction of p110d protein expression.
The presence of high p110d mRNA levels is not a consequence of leukocyte-specific p110d mRNA stability To assess whether high expression of the p110d protein in cells is due to increased mRNA stability, cells were treated with Actinomycin D, an inhibitor of de novo RNA synthesis, followed by measurement of mRNA decay over time. As can be seen from Figure 3A, leukocyte (A20 and EL4) and non-leukocyte (B16-BL6, 3LL and NIH-3T3) cell lines displayed very similar rates of mRNA degradation upon inhibition of mRNA synthesis, indicating that there is no difference in p110d mRNA stability between cell types expressing high or low levels of p110d protein. Also p110d protein levels were not affected by Actinomycin D treatment, both in leukocytes and non-leukocytes ( Figure 3B).

Identification of multiple distinct p110d mRNA transcripts with alternate first 59 untranslated exons
In order to identify the PIK3CD promoter, we set out to identify the transcriptional start site of the p110d mRNA. Rapid amplification of 59 cDNA ends (59RACE) was used to identify the 59UTR. BLAT alignment of the 59RACE products led to three main observations: (1) multiple distinct p110d transcripts exist within each cell line investigated; (2) most transcripts contains two untranslated exons ( Figure 4A), which we have named exon -1 and -2 (to indicate their relative locations with respect to exon 1, which contains the putative ATG translation start site as defined in [13,54]). The -1 and -2 exons are located 11 kb and .35 kb 59 of exon 1 in murine cells, and 19 kb and .59 kb 59 of exon 1 in human cells; (3) exon -1 can occur together with one of 4 identified second untranslated exons (exons -2a, -2b, -2c or -2d) in mouse cells, and with one of two -2 exons (-2a or -2b) in human cells. Thus, mouse and human PIK3CD can give rise to at least 4 and 2, respectively, distinct p110d transcripts.
Human and mouse p110d exon 1 contains an in-frame stop codon immediately upstream of the p110d start codon (for sequences, see Supporting Information, File S1), ruling out the possibility that the newly identified upstream exons are translated as part of the p110d protein. In other words, the -1 and -2 exons form the 59UTR of the PIK3CD transcripts.
The -1 exons identified in human and mouse show a high degree of homology (data not shown) and likewise the -2a exons share a region of high homology ( Figure 4B). Human exon -2a is approximately half the length of mouse exon -2a (72 bp compared to 144 bp), and its first part (1-49 bp) is homologous to the last part of mouse exon -2a (96-144 bp). Interestingly, the intronic DNA immediately upstream of human exon -2a is highly homologous with the first part of mouse exon -2a ( Figure 4B). In contrast the -2b exons for human and mouse are not homologous, which may indicate that the -2b exon identified in human may not be the equivalent -2b exon identified in mouse.
Database information provides independent confirmation of several of the untranslated exons identified in this study (-1, -2a, -2b and -2c in the mouse; and -1, -2a in human), as well as additional -2 exons in the mouse (labeled with an arrowhead in Figure 4B) which we have thus far not found by 59RACE both in mouse and human (.50 independent 59RACE products sequenced; detailed data not shown).
The p110d transcripts identified by 59RACE always contained exon -1, and further incorporated a single -2 exon in all cases. We have found two instances where the -1/-2 exon of the p110d mRNA configuration does not seem to occur. Firstly, during the cDNA cloning of human p110d [13], we identified one clone (called o5) which did not contain exon -1, and which has exon -2a directly spliced onto exon 1, giving rise to a p110d transcript that encodes full length p110d protein (data not shown). Secondly, 2 out of 48 transcripts identified by 59RACE in the mouse EL4 leukocyte cell line did not have -2 exons, and started with an exon -1. As shown in Table 1, all -2 and -1 exons contain a classical splice donor sequence (GT), whereas exon -1 only contains a splice acceptor sequence (AG). This is in line with the observation that each p110d transcript identified by 59RACE contains a single -2 exon and further indicates that the different p110d transcripts arise individually and not from a 'master' p110d transcript by alternative splicing. The presence of a splice acceptor sequence in exon -1 indicates that the p110d transcript starting at this exon (as revealed by 59RACE) found in mouse EL4 cells could in fact be an artefact, due to RNA degradation during the RACE experiments.
Cell type-specific usage of the multiple PIK3CD transcription start sites We next used RT-PCR to confirm the presence of the different p110d transcripts identified by 59RACE and to determine which of these can be found in a panel of murine leukocyte and nonleukocyte cell lines. Forward primers, specific for each of the 59 untranslated exons were designed, and used in combination with a common reverse primer in exon 2 (schematically shown in Figure 5A). PCR products of the predicted size were purified by agarose gel electrophoresis and verified by DNA sequencing (data not shown).
All cell lines tested (except CT26 colon carcinoma) expressed at least one type of transcript containing a -2 exon ( Figure 5B). Leukocytes contained a broader variety of p110d transcripts than non-leukocytes, with an average number of distinct p110d transcripts of 4.3 versus 2.4 in leukocytes and non-leukocytes, respectively (summarized in bottom panel of Figure 5B).
To more accurately quantify the amount of each p110d transcript, we next used real time RT-PCR ( Figure 6). For each transcript, the PCR reaction consisted of a forward and reverse primer, which were designed to amplify a cDNA sequence of ,100 bp spanning an exon boundary specific to the transcript, and a dye-emitting probe, which bound at a sequence overlaying this exon boundary. During amplification of the cDNA sequence, cleavage of a reporter dye from the probe results in fluorescence emission, which can be directly correlated with the level of each particular p110d transcript. For example, to measure the amount of p110d transcripts containing exon -2a, a PCR was performed using a forward primer in exon -2a and a reverse primer in exon -  1, while the dye-emitting probe bound at this exon-exon boundary. The fluorescence detected from this PCR reaction therefore represents the amount of all transcripts specifically containing exons -2a and -1.
These experiments revealed that leukocytes express significantly higher amounts of the different p110d transcripts than nonleukocytes ( Figure 6), indicating that leukocytes are likely to be more efficient at using p110d gene promoters than non-leukocytes.
In all cell lines, the transcript containing the first coding exon (exon 1; Figure 6A) was expressed at similar levels as the transcript containing the exon -2a/exon -1 boundary, which is the most abundantly expressed -2 exon ( Figure 6C).
Surprisingly, the transcripts containing the exon -1/exon 1 boundary ( Figure 6B) were two-fold more abundant than the transcripts containing the exon 1/exon 2 boundary ( Figure 6A). This indicates that shorter but still fully processed mRNAs (i.e. with a poly A tail since oligo d(T) was used for reverse transcription) are made. These would contain a 59UTR with at least the untranslated exon -1 and the coding exon 1, but without any of the other coding exons. Database analysis also provided evidence for such shorter p110d transcripts (marked with asterisks in Figure 4A). These may belong to the recently identified new class of mRNA transcripts that initiate near the expected transcription start sites, upstream of protein encoding sequences [55][56][57][58][59].
In silico analysis of PIK3CD promoter Alignment of the genomic sequence of flanking (and including) the 59UTR exons of mouse PIK3CD with 8 other species revealed high homology in specific areas, indicative for functionally conserved DNA sequences, including 4 CpG islands but no TATA boxes ( Figure 7A; Supporting information, File S2).
For each of the murine untranslated exons, the region spanning 500 bp upstream and 100 bp downstream of the first nucleotide were analysed for TF-binding sites and the transcription start site (TSS) prediction score within this region was assessed ( Figure 7B). TF-binding sites were identified in the vicinity of all mouse untranslated exons, however a particularly condensed cluster of TF-binding sites was identified within exon -2a ( Figure 7B, C). Interestingly, in human, this TF-binding cluster lies 59 of the TSS ( Figure 7C; schematically shown in Figure 7D). It is unusual, but not unheard of, that promoter regions are contained within exons. Indeed, recent work from the ENCODE project (http://www. genome.gov/10005107 and http://genome.cse.ucsc.edu/EN-CODE/) has revealed that proximal TF binding sites usually fall within 1 kb of both sides, 59 and 39, of the transcription start site [60].
The TF-binding cluster of murine exon -2a was located within a CpG island ( Figure 7A); was associated with a good TSS prediction score (0.9/1.0; Figure 7B) and was highly conserved across 28 species (Supporting information, File S2; Figure 7C shows the high degree homology of this region across 8 species, Figure 7D schematically shows the homology between human and mouse in this area). Collectively, these observations indicate the presence of a putative promoter region in/around exon -2a. Interestingly, 4 of the 7 different TFs identified within this binding cluster, namely ETS, IRF, NFAT and LEF (indicated by an asterisk in Figure 7C), have previously been associated with haematopoiesis and expression of leukocyte-specific genes (discussed in more detail below), suggesting that this TF-binding region may be involved in the high p110d expression in leukocytes.

Functional analysis of putative PIK3CD promoter elements using reporter assays
We next cloned intronic genomic DNA sequences that flank mouse exons 1, -2a and -2b (including exons -2c and -2d) at their 59 end (Figure 7A,B; referred to as DNA fragments A-I) as well as mouse exon -2a itself ( Figure 7B, DNA fragment J, which contains the TF-binding cluster), into the pGL3 reporter vector to drive expression of firefly luciferase. Vectors were transiently transfected in leukocyte and non-leukocyte cell lines and the promoter activities of the different PIK3CD DNA fragments were compared to that of the established leukocyte-specific promoter of Vav [61], and of the SV40 promoter, which is active in all cell types. The pGL3-Basic vector, which does not contain a promoter sequence upstream of firefly luciferase, was used to assess the basal level of luminescence.
The intronic genomic DNA fragments A-I did not possess significantly higher promoter activity in A20 leukocytes compared to NIH 3T3 fibroblasts ( Figure 8B). This is in contrast to DNA fragment J (mouse exon-2a containing the TF-binding cluster) which had significantly higher promoter activity in the mouse macrophage cell line RAW 264.7 than in NIH 3T3 fibroblasts ( Figure 8C), which was significantly higher than the leukocytespecific Vav promoter ( Figure 8C). The exon -2 fragment has also higher activity in the THP-1 monocytic cell line compared to the HEK293 (embryonic kidney) and CT26 (colon carcinoma) cell lines, again with higher activity compared to the Vav promoter ( Figure 8C). Taken together with the relatively high abundance of the -2a transcripts (Figure 6C) over the other p110d exon -2 transcripts ( Figure 6D-F), these data indicate that the TF binding cluster of exon -2a is the predominant promoter of p110d expression in leukocytes.

Discussion
In this study, we have explored the mechanisms by which the well-documented leukocyte-enriched expression of p110d might be achieved. We present evidence that p110d expression is mainly regulated at the transcriptional level arising from different transcripts. PIK3CD transcription appears to be constitutive and not responsive to acute cellular stimuli, at least using the agonists tested in the three cell types (NIH 3T3 and L929 fibroblasts, primary B cells and U937 myelomonocytic cells) used in our study. All cell types can express the distinct p110d mRNA transcripts but leukocytes express a greater diversity of transcripts and significantly higher amounts of the individual transcripts. In particular, the p110d transcript containing the untranslated exon -2a, was expressed at similar levels to the transcript containing coding exon 1 in a panel of murine leukocyte cell lines (compare Figure 6A with Figure 6C), indicating that the majority of p110d transcripts in a cell contain exon -2a.
We have identified a region within mouse exon -2a that contains a cluster of TF-binding sites. This TF-binding cluster is highly conserved between species and found immediately upstream of human exon -2a in human. This TF binding cluster contains at least 4 leukocyte-related TFs sites and was found to display higher promoter activity in leukocyte cell lines compared to nonleukocyte cell lines. Given that the majority of PIK3CD transcripts contain exon -2a, these data indicate that the promoter region identified within mouse exon -2a is likely to be sufficient in mediating the majority of leukocyte-specific PIK3CD gene expression.
4 of the 7 TFs identified within this binding cluster, namely ETS, IRF, NFAT and LEF have previously been associated with regulation of haematopoiesis and expression of leukocyte-specific genes. Indeed, the ETS family of TFs play important roles in the regulation of haematopoiesis [62][63][64][65][66][67][68]. IRF family members are highly expressed, but not exclusively, in cells of the immune system and play a pivotal role in the induction of type I IFN signalling pathways [69], proinflammatory cytokines and expression of macrophage and B cell specific genes [69,70]. NFAT family proteins are also mainly found in cells of the immune system, such as T cells, mast cells, NK cells and monocytes [71] and play a role in the regulation of various cytokines [72]. Finally, the LEF family of TFs, which are highly expressed in pre-B and T lymphocytes [73], have been linked to the regulation and expression of a number of lymphoid-specific genes [74][75][76][77][78][79].
High expression levels of p110d are also frequently observed in some non-leukocyte cancer cell lines, such as in breast carcinoma, melanoma and glioma [18]. It is possible that cancer cells upregulate or aberrantly express TFs which are, in non-cancer cells, more specific for leukocytes. It is of interest to note that a number of the TFs that bind in the exon -2a cluster have indeed been implicated in breast cancer progression, including LEF [80][81][82][83], ETS-1 [84,85], ETS-2 [86] and NFAT3 [87]. Recently, all four of these leukocyte-associated TF were identified as the most frequently differentially activated TFs in breast cancer based on a large microarray dataset [88].
We have found evidence that, among the multiple p110d transcripts, there may be mRNAs that do not encode full length p110d. Indeed, transcripts containing the exon -1/exon 1 boundary are more abundant than those covering the exon 1/ exon 2 boundary (compare Figure 6B to Figure 6A). Current database information supports the presence of such shorter p110d transcripts (marked with an asterisk in Figure 4A). Indeed, several recent studies have reported the discovery of a new class of short promoter-associated RNA transcripts that initiate near the expected transcription start sites upstream of protein-encoding sequences [56][57][58] (reviewed in [55]). It remains to be seen whether these RNAs have a function, but their prevalence suggests that their synthesis may serve a functional role.
Further work is required to understand the precise mechanism of p110d gene expression. The complexity of gene regulation has been exemplified by examination of 400 protein-coding genes in 1% (30 million bases) of the human genome as part of the ENCODE project [60], which revealed that 80% of these genes had additional exons, many of which were located thousands of bases away from the coding exons. Also many novel transcription start sites were found, many located thousands of bases away from the known start sites, while 25% of the promoters discovered were at the 39 end of the genes rather then at the 59 end. It is therefore highly likely that p110d expression will be subject to additional levels of control rather than by simple proximal promoter elements.
The data presented are the first to shed light onto the leukocyteenriched expression of PI3KCD. Further investigations are needed to identify which TF-binding sites are critical in driving PIK3CD gene expression and whether cells of non-leukocyte origin, such as breast cancer cells, are able to utilize this putative promoter.
Interference with PIK3CD expression at the promoter level may offer a novel therapeutic target in cases of aberrant p110d overexpression, as observed in some cancers [18].

RNA extraction, 59Rapid Amplification of cDNA Ends (59RACE), Reverse Transcription (RT)-Polymerase Chain Reaction (PCR) and Real Time RT-PCR
Total RNA was extracted from cells using the RNeasy mini kit (Qiagen, 74104). mRNA was subsequently reverse transcribed using SuperScript II Reverse Transcriptase (Invitrogen, 18064) and oligo d(T) primers, and subsequently used in 59RACE, RT-PCR or real time PCR.
For 59RACE, the FirstChoiceH RLM-RACE kit (Ambion, 1700) was used following the manufacturer's protocol using the following outer primers: murine 59-CAGATCAGCTCCTCA-TTGGCACT-39, human 59-GCTTCTTCACGCGGTCGCCC-39 and inner primers: murine 59-ACTTGAACTTCCCCGTG- For real time RT-PCR amplification TaqMan Universal PCR Mastermix (4304437) and primer mixes containing a FAM reporter probe (TagMan Gene Expression Assay) were obtained from Applied Biosystems. SYBR Green (Qiagen, 204143) was used for quantifying 18S RNA. Exon-specific primer sets and probes were designed to identify transcripts containing exon -1, -2a, -2c and -2d. For exon -2a the following primer sequences were used; forward primer 59-TCGCGCCTAGCCTTGG-39, reverse primer 59-GGCATCAGCGGGCTTCA-39 and FAM reporter sequence 59CTCAGCTCCTTAGATGTCGGTC-39. For exon -2b the following primer sequences were used; forward primer 59-AGTGTCTGTCCTGACTTCCTAAGAA-39, reverse primer 59-CGGGCTTCATCCCACTTCTG-39 and FAM reporter sequence 59-CAGCTCCTTAGATGTACTTCTACA-39. For each transcript of interest, known amounts of plasmids with this transcript were used to create a standard curve. Real-time PCR generated a series of C T values (the PCR cycle at which amplification of each target gene is first detected) for endogenous and plasmid-born cDNA, which allowed for

Western blot
Cells were lysed and immunoblotted for PI3K expression as described before [89]. Primary antibodies were detected using fluorescently-labeled species-specific secondary antibodies (antimouse IRDye 800-conjugated (Rockland) and anti-rabbit Alexa-Fluor 680-conjugated (Molecular Probes). Quantification was done using an Odyssey infrared scanner (LICOR) using the manufacturer's software. Signal intensities were normalized for an internal loading control such as b-actin or GAPDH.

Bioinformatic analysis of putative promoter elements in PIK3CD genes
The upstream sequences for the five untranslated exons of the murine p110d gene were inspected within the February 2006 (NCBI build 36) assembly of the Mouse genome using the UCSC genome browser [90]. Regions spanning 500 bp upstream and 100 bp downstream of the first nucleotide of each exon were analysed. The corresponding multiple species alignment was extracted using the Vertebrate Multiz Alignment & Conservation track [91] within the UCSC genome browser. The alignments were then screened for conserved TF binding sites using MatInspector [92] and a vertebrate factors subset of a of a proprietary database of Genomatix. In addition the candidate regions were inspected with Eponine [93], a probabilistic method for detecting transcription start sites, using a threshold of 0.9.

Reporter gene assays
PCR amplification of genomic DNA from C57Bl/6 mice was used to generate fragments for the reporter assays. The amplified PCR products were inserted into the pGL3 reporter vector (Promega). Transfections of NIH3T3 and A20 cells were performed using Qiagen Superfect or electroporation, respectively. Equal number of cells were washed and lysed, using Promega lysis buffer (to normalize for transfection efficiency) and then assayed for luciferase activity using the firefly luciferase substrate from Promega on the MicroBeta workstation (Perkin Elmer). The luciferase activity was normalized using a luciferase gene in a pGL3 reporter vector under the control of the SV40 promoter as well as a promoterless luciferase/pGL3 reporter vector. DNA of the lymphocyte-specific Vav promoter (construct HS21 mentioned in Ref. [61], which we cloned from the original b-galactosidase reporter construct into a pGL2 luciferase reporter vector) was used as a positive control.

Supporting Information
File S1 59 RACE product sequences Sequences of the different murine (Mm) and human (Hs) p110d transcripts as identified by 59RACE.