RNA splicing regulates the temporal order of TNF-induced gene expression

Abstract

When cells are induced to express inflammatory genes by treatment with TNF, the mRNAs for the induced genes appear in three distinct waves, defining gene groups I, II, and III, or early, intermediate, and late genes. To examine the basis for these different kinetic classes, we have developed a PCR-based procedure to distinguish pre-mRNAs from mRNAs. It shows that the three groups initiate transcription virtually simultaneously but that delays in splicing characterize groups II and III. We also examined the elongation times, concluding that pre-mRNA synthesis is coordinate but splicing differences directly regulate the timing of mRNA production.

In many gene induction systems, after an inducer is introduced, the mRNA products of different genes appear sequentially. This sequential appearance has generally been attributed to delayed transcription of some genes relative to others (1–3). It is thought that genes with compact chromatin or negative histone marks take some time to unravel or to acquire activating histone marks (4–8). However, it is possible that transcription is initiated nearly simultaneously at all induced genes, but processing of pre-mRNAs takes a variable time to be accomplished, leading to sequential appearance of mRNA. For that to be true, the time after inducer addition for genes to become available for transcription would have to be small relative to the time for processing the transcripts.

We have examined the causes of sequential gene transcription in one situation, TNF addition to cultured cells. This addition induces three waves of mRNA appearance: early production of group I mRNAs, slower production of group II RNAs, and very slow production of group III RNAs (9, 10). Most of the genes in these groups are induced by nuclear factor Kappa-light-chain-enhancer of activated B cells (NF-κB), but some, especially ones in group I, are induced by other pathways (9–12). We have previously shown that an important determinant of this temporal process of response to TNF is the half-lives of the mRNAs in each group. Group I RNAs are very short-lived whereas the others are longer-lived; many of the group III RNAs show no degradation over 24 h, indicating an indefinite lifetime (9). However, whereas the variable half-lives can explain some aspects of the temporal evolution following TNF addition, there must be other aspects to this sequential appearance of mRNAs. Here, we show that initiation of transcription following TNF addition is rapid and evident almost simultaneously at all induced genes, but the generation of mRNA is a more variable process. It appears that splicing of the pre-mRNA transcript is the rate-limiting process because the half-life of pre-mRNAs is the key variable.

Results

Distinguishing Pre-mRNA from mRNA.

For this study, it was necessary to use a methodology that would clearly distinguish pre-mRNA (the genomic transcript) from mRNA (the processed product). Wishing to do this by quantitative PCR (qPCR), we examined whether a judicious choice of PCR primers is capable of making the distinction between pre-mRNA and mRNA following TNF addition, as has been indicated in other settings (13, 14). Unspliced transcripts can be detected using PCR primers flanking an exon-intron junction (Fig. 1A). Spliced transcripts (mRNA) can be unambiguously detected with primers located in neighboring exons if the intervening intron is long (>500 bases) or spanning the junction of two exons if the intervening intron is short. The specificity of the PCR primers used in this study was tested by their qPCR melting curve and their production of specific single amplicons (Fig. S1A). The primers were chosen so that they will all amplify under the same conditions, and the linearity of the assays with the pre-mRNA primers was tested using dilutions of genomic DNA as template (Fig. 1B). This test also provided a proof of principle that the PCR-based technique could be used to compare the relative abundance of DNA (in particular, cDNA) molecules. Because most genes are represented twice per cell in the genome, the pre-mRNA primers detecting different coding regions should and did give the same relative copy number from the genomic DNA sample (Fig. 1B). The precision of the quantitation was further supported using different PCR primer pairs detecting the same spliced transcripts (Fig. S1B). All of the RNA samples were treated with DNase, and we confirmed that there was no detectable genomic DNA contamination before analysis by omission of the reverse transcription step for RNA detection.

Fig. 1.

Monitoring endogenous gene transcription induced by TNF using pre-mRNA kinetics. (A) A scheme of a PCR-based method for specific detection of unspliced and spliced transcripts following Singh and Padgett (13). Pre-mRNA can be detected by the primer pairs located either at the 5′ junction of the intron (e.g., J1) or 3′ junction (J2′) of the intron whereas the spliced transcripts (mRNA) is detected by the primer pairs located at different exons (e.g., M1) (the number indicating the specific intron analyzed). For comparison, noncoding regions (e.g., the promoter region, P1) are also analyzed from the same DNA or RNA samples. (B) Linearity and comparability test of PCR primers. Dilutions of mouse genomic DNA were prepared and quantitated by qPCR with different PCR primer pairs detecting different regions of the mouse genome as indicated at the right. (C) Total RNA isolated from mouse fibroblasts stimulated with TNF for the indicated time were first treated with DNase and then treated with (+) or without (−) reverse transcriptase (RT). PCR signals were measured by qPCR using primers specifically detecting the exon–intron junction 1 (J1) or promoter regions of Nfkbie. (D) The pre-mRNA dynamics of Fos and Tnfaip3 detected at indicated exon–intron junctions were monitored in mouse fibroblasts stimulated with TNF for indicated time using RT-qPCR. The results were average plus SD of duplicate samples and represent at least two independent experiments with similar results.

Response to TNF: Group I Genes.

We used this methodology to examine the induction of inflammatory genes in mouse fibroblasts and primary macrophages following addition of TNF. We chose five to seven highly inducible genes in each of the early (group I), intermediate (group II), and late (group III) gene families (9). The protein products of these genes have different biological functions, subcellular expression patterns, and basal mRNA expression levels (Table S1). Many studies have shown that RNA polymerase II (Pol II) initiates transcription frequently and widely, often generating abortive fragments that end 20–50 bases downstream of the start site (6, 15–17). To avoid amplifying these short transcripts, we have examined only exon–intron junctions at least 150 bases downstream of the transcription start sites and often much further away. We were able to detect the unspliced transcripts of all tested genes before TNF addition. The PCR signals from exon–intron junctions are much higher than those from nontranscribed regions (e.g., a promoter; Fig. 1C), indicating that the intron-containing RNAs detected here are not randomly initiated RNA fragments.

Group I genes are induced rapidly and repressed rapidly, producing a pulse of pre-mRNA that we would expect to easily detect by the PCR procedure. The genes we examined can be divided into NF-κB- dependent genes [e.g., chemokine C-X-C motif ligand 1 (Cxcl1), NF-kappaB inhibitor a (Nfkbia) (encoding IκBα), and tumor necrosis factor alpha-induced protein 3 (Tnfaip3) (encoding A20)] and NF-κB–independent genes [e.g., FBJ murine osteosarcoma viral oncogene homolog (Fos), dual specificity phosphatase 1 (Dusp1), and zinc finger protein 36 (Zfp36)]. Fos is activated by transcription factors Serum Response Factor (SRF) and Ternary Complex Factor (TCF) through the MAP kinase pathway and shut down by negative feedback mechanisms through dual specific protein phosphates (DUSP) family proteins (18–21). We first focused on Fos and Tnfaip3 to test whether pre-mRNA kinetics could detect differences in the kinetics of their transcriptional regulation. As expected, TNF induced a transient accumulation of unspliced transcripts from both Fos and Tnfaip3, but their kinetics were distinctly different. Fos transcripts were produced more rapidly and disappeared much more rapidly (Fig. 1D). PCR results generated using PCR primers located at different exon–intron junctions gave very similar results in terms of rise and decline times (orange and green line in Fig. 1D for fibroblasts and Fig. S1C for macrophages). These differential results for two genes induced by different transcription systems suggest that quite subtle aspects of pre-mRNA dynamics can be visualized by this relatively simple method.

We then compared the kinetics of unspliced and the corresponding spliced transcripts. Fig. 2 shows these kinetics for three genes, and more are displayed in Fig. S2A. As expected for RNAs that appear rapidly and disappear rapidly, the two curves have the same general form, except that the pre-mRNA is lower in abundance and its rise and decline occur a few minutes earlier than that of the mRNAs. The relative timing of the two curves is consistent with the precursor–product relationship of these two RNA classes.

Fig. 2.

The dynamic changes of unspliced (pre-mRNA in red) and spliced (mRNA in blue) transcripts were quantitatively compared between genes in three different groups by RT-qPCR in mouse fibroblasts stimulated with TNF for the indicated times. In each graph, pre-mRNA and mRNA are plotted together but with different scales; the left vertical axis is for pre-mRNA, whereas the right is for mRNA. The results are averages of duplicate samples with less than 15% variation and represent at least three different experiments with similar results. For larger time scale of the same results, see Fig. S2A. For the results in mouse bone marrow derived macrophages, see Fig. S2 B and C.

Group II and III Genes.

Surprisingly, the pre-mRNAs from group II genes rose with a kinetics quite different from that of their mRNA (Fig. 2, Center). The pre-mRNA kinetics arising from these genes were quite similar to those for the group I genes (Fig. 2 and Fig. S2A, red lines), especially for the group I NF-κB–dependent genes, consistent with previous studies that these group II genes are also regulated by NF-κB (10). The nearly simultaneous rise and fall of unspliced transcripts from these two groups of genes coincides precisely with the known kinetics of NF-κB nuclear entry and exit as described by us and others (22–24). The coincidence of NF-κB variation and that of the unspliced transcripts suggests that the timing of gene transcription is directly controlled by the intranuclear concentration of NF-κB but that mRNA production lags the transcription kinetics.

The mRNA production from the group II genes followed the pre-mRNA accumulation by a much larger lag than seen for the group I genes. Aside from the longer half-life of group II mRNAs as described elsewhere (9), the longer lag before appearance of spliced mRNAs was the primary difference between the two groups and strongly suggests that the mRNA kinetics are determined by the time for generating mRNA from pre-mRNA, the time for pre-mRNA processing. This processing is largely splicing out of introns and suggests that different groups of mRNA are produced at different speeds from pools of pre-mRNA that appear with very similar kinetics. In other words, splicing control, not transcription control, is responsible for the lagging appearance of group II RNAs.

The behavior of group III genes was particularly interesting. These genes have previously been considered inactive at the early stage of induction (at least for the first half hour) following TNF addition (9, 23). As opposed to the group I and II genes, which are transcribed by factors rapidly mobilized to the genes (primary genes), group III genes are divided into a major group of secondary response genes and a minor group of primary response genes (4, 9, 25). The secondary response genes require protein synthesis to induce their later transcription, suggesting that induced transcription factors are involved in this transcription (3). Most of the group III genes are NF-κB–dependent for their transcription. Surprisingly, we detected significant early transcriptional increases of the pre-mRNAs of all five group III genes we tested (Fig. 2, Right and Fig. S2A), no matter whether they were primary response genes [e.g., chemokine C-C motif ligand 5 (Ccl5)] or secondary response genes [e.g., matrix metallopeptidase 3 (Mmp3) and serum amyloid A3 (Saa3)]; for the list of primary and secondary genes, see refs. 4 and 26. This clearly indicates that the group III genes are activated much earlier than had been evident from their mRNA accumulation kinetics. It is also evident that late genes are different from the early and intermediate genes in that their transcription is less attenuated, or even enhanced after the first wave of transcription, indicating that late genes are ones that are further activated at a later stage of the TNF response.

The strong similarity among the groups in pre-mRNA kinetics (Fig. 2, red lines), in sharp contrast to their different mRNA accumulation kinetics (Fig. 2, blue lines), reinforces their similarity in transcriptional regulation and explains why, in our earlier work (9), swapping promoters between early and late genes did not result in corresponding changes of mRNA induction patterns. This similarity emphasizes the important role played by differences in mRNA stability in determining the distinct mRNA accumulation kinetics of the three groups (9).

Similar pre-mRNA kinetics between groups in response to TNF were also found in mouse primary bone marrow-derived macrophages (BMDM) (Fig. S2B; and for longer times, see Fig. S2C). This similarity shows that oscillatory transcription induced by continuous TNF stimulation (as the result of the oscillation of NF-κB into and out of the nucleus) is as true for hematopoietic as it is for fibroblastic cells and is likely to be a general phenomenon in all cell types. Its physiological significance remains to be established.

More Precise Timing.

Having found that all of the genes we examined began transcription within the first 10 min following induction with TNF, we wished to determine this time more precisely and thus took samples very frequently during this early time. We found that all genes began transcription within the first 9 min (Fig. 3A). We quantitated the precise rise times of pre-mRNAs [defined as the time after which the pre-mRNAs are continuously higher (P < 0.05 by t test) than the basal level] for genes in the three groups and present these values in Table S2. There is a slight tendency for the pre-mRNAs of group I genes to rise earlier than that of group II and III genes, but transcription of all genes rose within a 4-min window.

Fig. 3.

The temporal order of gene expression is not primarily set at transcription initiation. (A) The changes of pre-mRNA (A) and mRNA (B) of genes from group I (red), II (green), and III (blue) in mouse marcophages (BMDM) upon TNF stimulation were measured by RT-qPCR and shown as fold induction over their basal levels. The results are average values of duplicate samples with less than 15% variation and represent three independent experiments with similar results. For results with mouse fibroblasts, see Fig. S3. C3, complement component 3; Gbp2, guanylate binding protein 2; Il1b, interleukin 1, beta; P100, nuclear factor of kappa light polypeptide gene enhancer in B-cells 2; Relb, v-rel reticuloendotheliosis viral oncogene homolog B; Slc2a6, solute carrier family 2, member 6; Tlr2, toll-like receptor 2.

We compare the mRNA rise with the pre-mRNA rise in Fig. 3A. For group I genes, the lag is again seen to be short, but now the different behavior of most group II and group III genes becomes quite evident. The mRNA for these latter genes only begins to rise after about 20 min and for some group III genes it is even later (Fig. 3B). Similar results were found in mouse fibroblasts (Fig. S3 A and B).

To examine whether the temporal patterns of mRNA kinetics are mirrored in the order of protein expression and to extend the analysis beyond the mouse cell types described above, we examined human primary fibroblasts with a group II gene, CCL2, and a group III gene, CCL5, as examples. As expected, we found a sequential rise of the mRNAs for CCL2 (group II) and CCL5 (group III) (Fig. S3C, Left), with accumulation kinetics very similar to that found previously for group II and III genes in mouse fibroblasts and macrophages. Protein accumulated with very similar kinetics (Fig. S3C, Right) but with an even longer time lag following induction. Thus, as one would expect, the mRNA kinetics determine protein synthesis kinetics, with the products of group III mRNAs taking hours before they accumulate, although then accumulation continues indefinitely because of the long half-life of the group III mRNAs (9).

Time for Splicing.

If the slower accumulation of mRNA compared with pre-mRNA is due to splicing, then the conversion of pre-mRNA to mRNA should not be affected by the addition of actinomycin D (ActD), an inhibitor of transcription. Thus, we allowed TNF-induced pre-mRNA to accumulate for 15 min and then blocked further transcription with ActD. For group I pre-mRNAs, there was a very fast decline of pre-mRNA, with many of them having only 10% or even less left after 4–6 min. A much slower decrease of pre-mRNA was seen for the group II and III pre-mRNAs (Fig. 4A, Fig. S4A). This relatively slow decrease was also true in cells pretreated with TNF for 6 h (Fig. S4B), when transcription of early genes has been greatly decreased, whereas transcription of many intermediate and late genes is still high or even further enhanced (Fig. 2B and Fig. S2E). Pre-mRNA half-life was calculated based on the rate of decline of pre-mRNA after treatment of cells with ActD, showing that the pre-mRNA half-life is very short for early (group I) genes and longer for intermediate (group II) genes and late (group III) genes (Table S3). These relative half-lives were found in fibroblasts too (Fig. S4 C and D and Table S4). The fact that these differences are evident in ActD-treated cells strongly suggests that the pre-mRNA half-lives are not determined by gene transcriptional activity and are likely an intrinsic property of the completed transcripts.

Fig. 4.

RNA splicing also regulates the temporal order of gene expression. (A) Mouse macrophages (BMDM) were first pretreated with TNF for 15 min before the addition of the transcription inhibitor actinomycin D (ActD). The average pre-mRNA fold change (in log scale) after ActD addition of group I (red), group II (green), and group III (blue) is shown in A. For the results detected at specific locations by RT-qPCR, see Fig. S4A. (B) In the same experimental setting, the changes of unspliced (pre-mRNAs in red) and the corresponding spliced transcripts (mRNAs in blue) of a group I gene, Tnfaip3 (Upper), and a group II gene, Icam1 (Lower), after ActD addition were measured by RT-qPCR. The results are average plus SD of relative copies and represent two independent experiments with similar results.

To confirm that the decline of pre-mRNA reflects RNA splicing timing rather than nonspecific nuclear degradation, we measured the changes of unspliced and corresponding spliced transcripts from the same samples after addition of ActD. We found that the decrease of unspliced transcripts for group I genes (e.g., Tnfaip3; Fig. 4B, Upper) and group II genes [e.g., intercellular adhesion molecule 1 (Icam1); Fig. 4B, Lower] was accompanied by an increase of the corresponding spliced form of the transcript and that the amount lost in pre-mRNA roughly matched the amount gained in mRNA (Fig. 4B). This equality strongly suggests that the majority of the unspliced transcripts were successfully spliced in the presence of ActD.

There were some differences in pre-mRNA half-life measured at different introns of individual genes (Table S3), possibly suggesting that introns may be spliced independently, but this observation requires further study. Pre-mRNA in general have longer half-life in mouse fibroblasts than in mouse macrophage (comparing Tables S3 and S4). We saw a similar phenomenon for the mRNA half-life between these two types of cells (9). However, in all measurements, the differences between early genes and the other two groups remained evident (Tables S3 and S4, Fig. 4A, and Fig. S4 A–D). This consistency among cell types suggests that the differences in RNA splicing reflect a gene intrinsic property. Thus, intrinsically longer RNA splicing times appear the best explanation for the delayed mRNA appearance for group II and III genes.

RNA Splicing Is the Rate-Limiting Step for mRNA Production.

One possibility for the delayed appearance of mRNA from one gene compared with another would be a slower transcription rate. Thus, we examined the synthesis rate for two genes: group I gene Tnfaip3 and group II gene Icam1. Compared with other group I genes, they are relatively long (14 kb and 13 kb, respectively) and allow particularly accurate measurement of the elongation speed. The elongation speed was estimated from the time for the transcription wave induced by TNF to travel between two exon–intron junctions. This is an unperturbed form for measurement of the Pol II transcription speed on endogenous genes in live cells under physiological conditions. We saw an ∼1.6-min lag between the rise of pre-mRNA at exon 2 (J2) of Tnfaip3 and at exon 8 (J8) (a distance of 8.1 kb) (Fig. 5A). These values translate into an average elongation speed of 5 kb/min, which is somewhat faster than reported elsewhere based on in vitro or artificial reporter systems (13, 27–29). A similar high elongation speed (5.2 kb/min) was found for group II gene Icam1, between the 5′ end (J2) and 3′ end (J2′) of intron 2 in macrophages (Fig. 5B). This similarity in elongation speeds clearly indicates that the delayed appearance of Icam1 mRNA is not due to slowness of transcription, at least over the regions that we studied. A similar result is also observed in mouse fibroblasts, in which an even higher elongation speed (∼8 kb/min) was reproducibly observed for both A20 and Icam1 (Fig. S5A). We have also examined other genes and, although elongation speed appears to vary somewhat from region to region of DNA (Fig. S5B), we reach the same general conclusion: splicing of a transcript, even at introns of group I genes, is a slow process relative to transcription.

Fig. 5.

RNA splicing is a relatively slow process compared with transcription. (A and B) The scheme of the structures of a group I gene, Tnfaip3 (A), and a group II gene, Icam1 (B), with the location of two detecting PCR amplicons is shown in Upper (not in scale). (Lower) The pre-mRNA changes detected at these two regions at indicated time points were measured by RT-qPCR. The data points are average plus SD of duplicate samples and represent two independent experiments with similar results. The time lags were estimated from linear regression analysis (the lines) of initial pre-mRNA kinetics detected at the two regions.

Discussion

In this study, we attempted to illuminate a crucial aspect of gene induction: Is there gene-specific variability in the time it takes for a transcript to be processed into mRNA? We were led to this question by the observation that mRNA appears in three waves following TNF addition (9), implying that some step(s) in mRNA appearance must differ for different gene transcripts. We have previously shown that mRNAs differ in their half-lives, with the rapidly appearing mRNAs having the fastest degradation times (9). This observation explained some characteristics of the different waves of mRNA appearance but by no means all. Thus, we decided to measure the kinetics of pre-mRNA synthesis and compare them to the kinetics of mature mRNA synthesis in an intact physiological setting. To make these measurements, we adapted PCR methodology previously used for very long genes (13, 29). We found that the kinetics of pre-mRNA synthesis could be accurately measured in this way and compared with mRNA kinetics. The actual measurements were of the amounts of RNA that were either unspliced or spliced at a given splicing junction, but we studied multiple junctions and multiple RNAs, getting quite consistent pictures.

The data showed that, for the group I (early) genes, there was only a few minutes lag between synthesis of a pre-mRNA and its appearance as a spliced mRNA. That would be expected because these RNAs appear rapidly in a pulse and are rapidly degraded. However, for group II and group III genes, the lag between synthesis and of pre-mRNA and its splicing was much longer, making splicing rate the controlling factor for mRNA appearance rate. These studies support a model where the temporal order of gene expression is not primarily determined by the timing of the initiation of transcription but by gene intrinsic properties that are determined by DNA sequences in and around the coding regions of the genes (e.g., promoters, introns, and 3′ UTRs).

Sequential transcriptional initiation has been considered to be the main step regulating the order of gene expression in many systems, and the evidence for this assertion is strong, for instance, in early Drosophila development (1–3, 28). Surprisingly, we found in a nondevelopmental process of sequential gene induction that, following TNF treatment of various cell types, pre-mRNAs of the early (group I), intermediate (group II), and late (group III) genes all rose rapidly within a very narrow time window. Many steps ensue between the addition of TNF to culture media and the increased rate of appearance of mRNAs from particular genes. These steps include binding of the TNF to its receptor, intracellular signal transduction, activation of latent transcription factors (here mainly NF-κB), movement of NF-κB into the nucleus, clearance of impediments to binding of NF-κB to DNA, binding of other factors and possible histone modifications, and the activation of RNA polymerase II (pol II) to polymerize the pre-mRNA (30). Our experiments show that the minimal time for the overall response to TNF is more than 4 and less than 9 min at all genes measured but do not know how to apportion these few minutes among the many processes occurring. Also, all of the genes we have studied have basal rates of transcription in the absence of inducer, so we are not measuring de novo activation of the genes but rather increases of transcription rates. Clearly, the temporal order of the response of genes to TNF induction is not primarily set at the transcription initiation stage for the genes we have studied.

We chose to study this question using PCR methodology. The issues raised here could in principle be addressed using genome-wide analytic techniques like RNA-seq or microarrays. However, these global techniques do not as precisely and sensitively separate pre-mRNA and mRNA or have the inherent linearity provided by qPCR that allows quantitative comparison of these two types of RNA. It would also be difficult to get the time resolution we achieved. However, adaptation of these global techniques to the questions raised by this study should be valuable.

Previous studies have indicated that the elongation rate of mammalian RNA polymerase II is about 3–4 kb/min [reviewed by Darnell (31)]. We found a somewhat faster rate, perhaps because we used different cells or different techniques. The method we used does not involve any perturbation of the cell by drugs whereas previous studies have often used cells recovering from inhibition by the drug DRB (5,6-dichloro-1-b-d-ribofuranosylbenzimidazole). In any case, our estimates are not materially different from previous ones—the key point for this study is that pre-mRNA synthesis can be completed in a much shorter time than splicing is completed.

Our observation that introns are removed from pre-mRNA at different rates is consistent with a body of past data on viral and Drosophila genes [reviewed by Darnell (31)]. We do not know what aspects of pre-mRNA metabolism might control the timing. We did note that often multiple sites in one pre-mRNA are spliced at similar rates, but we were unable to examine all of the splice sites in individual pre-mRNAs to determine the total range of speeds among sites in one pre-mRNA.

In general, it appears that RNA splicing is a process that can be dissociated from transcription. With a rate of 5 kb/min for transcription but splicing happening over many minutes, the two processes appear quite separable, at least in this system. Thus, a pool of unspliced or partially spliced pre-mRNAs can accumulate following TNF stimulation, especially for the group II and III genes. In this way, multiple introns can apparently coexist in the same nascent transcripts, providing a substrate for possible alternative splicing and other editing processes. Also, the timing of mRNA production is heavily dependent on the speed of RNA splicing; thus, the differences in RNA splicing rate become an important mechanism for regulating the temporal order of gene expression following a single induction stimulus.

Rabani et al. (32) addressed these same questions using an RNA metabolic labeling approach but reached quite different conclusions from those in our previous and present studies; they suggest that transcription rate determines the temporal changes of mRNA levels. The discrepancy may derive from two considerations. First, their study used a metabolic labeling approach, which is limited for study of fast processes like transcription and RNA splicing because the sensitivity of detection depends on the time of labeling. Even though their labeling time was as low as 10 min, that is still much longer than the half-lives of many pre-mRNA (happening in minutes; Tables S3 and S4). Therefore, the label so rapidly accumulates in newly spliced RNA (mRNA) that the pre-mRNA labeling is obscured. The rapidity of splicing may explain why their newly labeled RNA kinetics is so similar to that of mRNAs.

Recently, Bhatt et al. (33) analyzed pre-mRNA and mRNA kinetics in macrophages stimulated by lipid A (the active component of LPS) using subcellular fractionation and RNA deep sequencing techniques. Their primary conclusion—that following induction there is a rapid accumulation of unspliced chromatin-associated transcripts that behave like precursors of spliced mRNA—is in strong agreement with our conclusion that splicing is a slow step in mRNA maturation. They could not resolve differences in rates of splicing at different genes. They also concluded that transcription is the primary regulator of the temporal induction of inflammatory genes, a conclusion different from that in our previous and current studies. Various differences in methodology may account for this discrepancy: (i) We studied TNF induction; their study of lipid A induction is a more complex situation because LPS activates more pathways than TNF, producing multiple mediators, such as TNF, IL-1b, IFN-b, IL-10, and changing mRNA stabilities (9, 25, 34, 35). These factors can produce secondary effects. (ii) Their study analyzed transcriptional kinetics using a limited number of time points (five time points in a 2-h poststimulation window) and averaged cellular responses of groups of genes, which vary greatly in fold-induction levels. As a result, the rapid changes of individual pre-mRNAs may have been missed and/or smoothed out. Interestingly, the one result from an individual gene that they show, that for nuclear factor kappa-light-chain-enhancer of activated B cells (Nfkb1), differs from the average kinetics but is quite similar to our patterns. We believe that their study is not in serious contradiction to the conclusions we have made and is strongly supportive of the relatively slow nature of splicing that we describe.

Pre-mRNA splicing has previously been shown to be important for the regulation of nuclear export and the efficiency of protein synthesis (36, 37). Here, we show that pre-mRNA splicing can also regulate the timing of mRNA appearance from activated genes. What we have seen may be mechanistically related to the widely observed phenomenon known as intron delay, in which an intron serves as an important timing mechanism for early animal development (38, 39). The present study and our previous study suggest that the pattern of expression of a gene is determined not just by transcription, but is also importantly controlled by RNA splicing and mRNA degradation. Thus, multiple layers of gene regulation work in concert to generate a diverse set of gene programs while using only a limited number of transcription factors.

Materials and Methods

Cell Culture and Pre-mRNA and mRNA Analysis.

Mouse embryonic immortal 3T3 fibroblast cell lines, human primary dermal fibroblasts, and mouse bone-marrow derived macrophages were generated or purchased, maintained, and stimulated with TNF as previously reported (10). Briefly, cells were stimulated with TNF for indicated times and lysed for isolation of RNA using the RNeasy Mini kit (Qiagen). Each experimental condition was performed in duplicate or triplicate. For quantitative analysis of pre-mRNA and mRNA, 1 μg of total RNA was first treated with DNase, split into two parts, and either treated with or without reverse transcriptase using the Invitrogen SuperScript III first-strand synthesis kit. qPCR was performed using Sybr Green PCR Master Mix (Kapa Biosystems) and ABI7300 Real-Time PCR machine (Applied Biosystems Inc) with specific PCR primers. The samples without reverse transcriptase treatment provided negative controls. The expression profiles of the pre-mRNAs and mRNAs of target genes were analyzed with specific PCR primers as shown in Fig. 1A from the same cDNA samples. The results were normalized to the mRNA amount of the housekeeping gene L32 (2⁻¹² of the L32 mRNA defined as 1 unit of activity). The sequences of the PCR primers are available in Dataset S1.

Pre-mRNA Half-Life.

Actinomycin D (ActD) (10 μg/mL) was directly added to cell cultures that were already treated with TNF for the indicated times without removing the original stimulant. The cells were harvested right before addition of ActD (time 0) or 4, 10, and 30 min after addition of ActD. Each condition was performed in triplicate samples. The amounts of pre-mRNA were measured by quantitative RT-PCR (qRT-PCR) as described above and normalized to the mRNA level of L32 before calculation of half-lives. The half-life (T) was calculated from the slope of pre-mRNA concentrations after addition of ActD. For those pre-mRNAs with very short half-lives (e.g., group I genes and Actin-b), we used , where Ct(t₁) and Ct(t₂) are the Ct values at time t₁ (0 min) and t₂ (4 min) after addition of ActD. For pre-mRNAs with longer half-lives, the slope was calculated based on linear regression analysis of three time points (4, 10, and 30 min) using Microsoft Excel.

ELISA.

Supernatants were collected from cell cultures at various times after TNF stimulation. The antibodies and recombinant proteins were from R&D Systems. ELISAs were done using R&D Systems ELISA kit as recommended by the manufacturer.