A negative elongation factor for human RNA polymerase II inhibits the anti-arrest transcript-cleavage factor TFIIS

Abstract

Formation of productive transcription complexes after promoter escape by RNA polymerase II is a major event in eukaryotic gene regulation. Both negative and positive factors control this step. The principal negative elongation factor (NELF) contains four polypeptides and requires for activity the two-polypeptide 5,6-dichloro-1-β-d-ribobenzimidazole-sensitivity inducing factor (DSIF). DSIF/NELF inhibits early transcript elongation until it is counteracted by the positive elongation factor P-TEFb. We report a previously undescribed activity of DSIF/NELF, namely inhibition of the transcript cleavage factor TFIIS. These two activities of DSIF/NELF appear to be mechanistically distinct. Inhibition of nucleotide addition requires ≥18 nt of nascent RNA, whereas inhibition of TFIIS occurs at all transcript lengths. Because TFIIS promotes escape from promoter-proximal pauses by stimulating cleavage of back-tracked nascent RNA, TFIIS inhibition may help DSIF/NELF negatively regulate productive transcription.

Regulation of productive mRNA chain synthesis by RNA polymerase II (RNAPII) occurs when the RNA chain is ≈10–100 nt long and determines whether RNAPII forms a fully functional transcription elongation complex (TEC) or halts RNA synthesis during the early stages of transcript elongation (reviewed in refs. 1–5). Conversion to a productive TEC involves an ordered set of transitions that require successive phosphorylation of Ser-5 and Ser-2 in the RPB1 C-terminal heptapeptide repeat domain (CTD) by kinase components of TFIIH and positive elongation factor P-TEFb, respectively. These CTD changes orchestrate release of initiation factors, recruitment of general elongation and chromatin-modifying factors, and capping of the nascent transcript by the capping enzyme. Completion of these steps is required to produce a TEC able to transcribe through nucleosomes on a chromatin template and synthesize full-length pre-mRNAs.

During this process, negative elongation factor (NELF) is recruited to nascent TECs through interactions with the 5,6-dichloro-1-β-d-ribobenzimidazole-sensitivity-inducing factor (DSIF) and RNAPII. DSIF is a heterodimer of SPT4 and SPT5 that alone can stimulate transcript elongation in some conditions (6, 7). NELF is a complex of four subunits: the Wolf–Hirschhorn syndrome candidate protein 2 (NELF-A), a cofactor of BRCA1 (NELF-B), one of two TH1-like products of alternatively spliced mRNAs (NELF-C/D), and an RNA-binding protein, RD (NELF-E). A region of NELF-A that resembles hepatitis ∂ antigen binds RNAPII (8), whereas a C-terminal region of RD (NELF-E) binds RNA (9) and is proposed to mediate inhibition of transcription through interaction with nascent RNA as the RNA emerges from the RNAPII exit channel (8). Both SPT5 and RD (NELF-E) also can be phosphorylated by P-TEFb (10–12), which may contribute to the formation of productive TECs.

Although DSIF/NELF interacts with RNAPII and RNA and slows or arrests extension of transcripts in vitro (13, 14), neither the fundamental mechanism by which it inhibits nucleotide addition nor the regulatory role of DSIF/NELF inhibition of RNAPII is fully understood. NELF-induced slowing or arrest of nascent TECs may both provide time for productive TEC-generating steps like mRNA capping and serve as a checkpoint to block RNA synthesis in the absence of regulatory input (12, 15, 16). For instance, DSIF/NELF induces promoter-proximal pausing in the Drosophila hsp70 gene (15), which is relieved upon heat shock through interactions of a heat-shock factor with the TEC (17). To allow productive transcription, NELF may be inactivated or released when P-TEFb phosphorylates SPT5, RD (NELF-E), and the RNAPII CTD (12, 14, 18); capping enzyme, which binds DSIF and the CTD, may help inactivate NELF (19).

The transcript cleavage factor TFIIS is another participant in early elongation. TFIIS promotes productive elongation by stimulating cleavage of backtracked transcripts that arise as RNAPII enters and resides in pause and arrest states during early elongation, including promoter-proximal pauses in the Drosophila hsp70 gene (20, 21). Similarly, the bacterial factor GreA is required for the escape of RNAP from promoter-proximal pauses upon recruitment of the positive elongation factor λQ (22). Both TFIIS and GreA stimulate cleavage of backtracked RNAs by inserting in the RNAP secondary channel and stabilizing the binding of the second catalytic Mg²⁺ ion in the RNAP active site (Mg²⁺II; refs. 23–26). Thus, we wondered whether the inhibitory effect of DSIF/NELF on productive elongation might in part involve counteracting the stimulatory effect of TFIIS on formation of productive TECs. To investigate this possibility, we examined the effects of DSIF/NELF and TFIIS separately and together on TECs halted in the early transcribed region of HIV-1 in vitro.

Materials and Methods

Template DNA. DNA templates with upstream biotin tags were prepared by PCR amplification of plasmid pLL283 or pPM151A (truncated HIV-1 LTR) and immobilized on streptavidin-agarose beads (Sigma), as described (27, 28). pPM151A was derived from pLL283 by deleting DNA coding for A15 through G48 of the HIV-1 transcript (template 2; Fig. 1A), yielding a truncated transcript with the pause site located at position +U28.

Fig. 1.

Comparison of effects of DSIF/NELF on TFIIS-stimulated transcript cleavage and on nucleotide addition. (A) Transcription templates. Sequences are shown for template 1 (wild-type HIV-1 transcript with the pause site indicated at U62) and template 2 (truncated HIV-1 transcript lacking anti-TAR and TAR with the pause site indicated at U28). Anti-TAR, dotted overline; TAR, black underline; truncation, open box. (B) Halted C27, U28, and U14 complexes (≈10 pM) were incubated with DSIF (≈60 nM) and NELF (≈7 nM) for 2 min at 30°C before addition of TFIIS (≈50 nM). Only a fraction of DSIF was active (14). Samples were removed at 8-s intervals and separated on a 12.5% denaturing gel (see Materials and Methods). The plots below each gel show the disappearance of C27, U28, and U14 RNA (•, +DSIF/NELF; ○, –DSIF/NELF). (C) TECs halted at U11–U14 on template 1 were elongated with all four NTPs (1 mM each). Aliquots were removed at 7-s intervals; a final sample (lane C) was removed 4 min after raising GTP to 5 mM. RNA products were analyzed as described (28), and the fraction U14 RNA with error analysis from three replicate experiments was plotted against reaction time. The entire gel is shown in Fig. 5, which is published as supporting information on the PNAS web site. (D) TECs halted at A26 formed on template 2 were elongated through the pause site in the presence of CTP, GTP, and UTP. Samples were collected, processed, and analyzed as described for C. (E) Ratio of pause strengths (product of pause duration and pause efficiency; refs. 30 and 31) at different positions on template 1 in the presence and absence of DSIF/NELF represented as a bar graph. Pause strengths correspond to the areas under the curves in C and D and were calculated (Materials and Methods) for full time courses and RNA electrophoretograms, an example of which is shown in Fig. 5.

Transcription Factors. Human transcription factor TFIIS was overexpressed in Escherichia coli and purified as described (29). DSIF and NELF were purified as described (14).

Transcription Reactions and Pause Strength Calculations. U14 TECs were prepared, converted to TECs with longer RNAs, and used for transcription assays as described. (27). Pause strengths (Fig. 1E) were calculated by sampling RNAs present at time intervals after addition of 1 mM NTPs to a set of initially formed TECs halted and washed at and before U14 (30). We prepared plots of the relative amounts of RNA vs. time (minus RNAs that had not reached the site in the final time point; plots similar to those in Fig. 1 C and D). Relative pause strength was calculated as the area under the line for each RNA species in the presence of DSIF/NELF divided by the area under the line in the absence of DSIF/NELF.

Transcript Cleavage Reactions. TECs halted at U14, U16, C27, U28, or U62 were incubated with TFIIS in the absence or presence of DSIF/NELF at concentrations and for times indicated in the figure legends. Samples were processed, separated, and quantitated as described (28). Mn²⁺-induced intrinsic transcript cleavage was performed on U28 TECs that were washed with transcription buffer containing MnCl₂ (10 mM) instead of MgCl₂. Pyrophosphorolysis was performed by incubating U28 TECs with PPi (1 mM) for 5 min at 30°C.

Results

DSIF/NELF Exhibits Two Independent Activities: Inhibition of Nucleotide Addition and of TFIIS. To test for an effect of DSIF/NELF on TFIIS action, we prepared halted TECs on two HIV-1-derived templates and then compared effects on TFIIS-mediated transcript cleavage with effects on transcript elongation at the same template positions. (TFIIS stimulates transcript cleavage in many halted or paused TECs but with release of smaller 3′ RNA fragments than from arrested TECs; ref. 3.) One template contained the promoter and early transcribed region of HIV-1, including a well characterized pause site at U62 known to form TECs susceptible to TFIIS-stimulated transcript cleavage (template 1, Fig. 1 A; refs. 27 and 28); the other (template 2) contained a deletion of sequences coding for the TAR (Transactivation responsive) and anti-TAR RNA structures. These RNA structures influence, but are not required for, pausing (28). Thus, paused TECs still form at U28 of template 2.

DSIF/NELF inhibited TFIIS-stimulated transcript cleavage in different halted TECs (C27 and U28 on template 2; U14 on template 1; Fig. 1B). We assayed TFIIS using halted TECs to avoid complications of simultaneous nucleotide addition and transcript cleavage but verified that transcript-cleaved TECs remained elongation-competent with or without DSIF/NELF when NTPs were added (data not shown). We observed similar DSIF/NELF inhibition of TFIIS cleavage with G12, U16, and U62 TECs on template 1 or 2 (data not shown). Although the rate of TFIIS cleavage varied among TECs (from ≈0.02 s^–1 for U28 TEC to ≈0.075 s^–1 for C27 TEC at ≈50 nM TFIIS), the magnitude of the DSIF/NELF inhibitory effect was relatively constant for all TECs (3.4 ± 0.15-fold at ≈7 nM DSIF/NELF). This result is consistent with DSIF/NELF inhibiting by constant amounts in either TFIIS binding to RNAPII or TFIIS stimulation of cleavage, whereas the basal rate of TFIIS-stimulated cleavage is determined by the RNA sequence in the active site and the extent of TEC backtracking, which vary among TECs.

In contrast to the constant effect of DSIF/NELF on TFIIS-stimulated transcript cleavage, DSIF/NELF exhibited variable effects on the rate of nucleotide addition. DSIF/NELF actually accelerated nucleotide addition at U14, whereas it inhibited nucleotide addition at both C27 and U28 (Figs. 1 C and D and 5). This result is consistent with the idea that DSIF/NELF inhibition of nucleotide addition involves interaction of RD (NELF-E) with nascent RNA (8, 9), because the U14 RNA should be entirely sequestered within RNAPII and unavailable for interaction with DSIF/NELF. Previous reports suggest inhibition by DSIF/NELF requires a minimum length of nascent RNA (13, 14, 19). We conclude that DSIF/NELF possesses two different inhibitory modes. One mode, described previously, inhibits nucleotide addition when nascent RNA is approximately ≥18 nt. The second mode, described here, inhibits TFIIS-stimulated transcript cleavage irrespective of transcript length.

To ask at which transcript length DSIF/NELF begins inhibiting nucleotide addition, we examined the effect of DSIF/NELF on transcript elongation at all positions after G11 on template 1. We calculated the apparent pause strength at different template positions and then plotted the ratio of apparent pause strengths with and without DSIF/NELF as a function of template position (Fig. 1E; see Materials and Methods). Pause strength is the product of pause duration multiplied by pause efficiency but is simpler to measure than pause duration or efficiency individually (30, 31). When the nascent transcript was 17 nt or shorter, DSIF/NELF reduced relative pause strength (i.e., increased the rate of transcription), whereas when the transcript was 18 nt or longer, DSIF/NELF increased relative pause strength (i.e., decreased the rate of transcription). This result suggests that DSIF/NELF commences inhibition when the RNA chain is 18 nt, which is roughly the position at which nascent RNA emerges from RNAPII (see Discussion). DSIF/NELF acceleration of RNAPII when RNA is ≤17 nt may be caused by the previously reported stimulatory activity of DSIF (6).

Both DSIF and NELF Are Required to Inhibit TFIIS. Enhancement of pausing by DSIF/NELF requires both the DSIF and NELF components (14). To ask whether inhibition of TFIIS also depended on both NELF and DSIF, we tested them alone and in combination. Even at the same concentrations that inhibited TFIIS when combined, DSIF and NELF had no effect on TFIIS-stimulated cleavage when added individually to U28 TECs (Fig. 2A). Thus, like the stimulation of pausing, the inhibition of TFIIS by DSIF/NELF requires both factors. Importantly, this result also demonstrates that TFIIS inhibition is a functional consequence of DSIF/NELF interaction and not an adventitious effect of a contaminant in DSIF or NELF preparations.

Fig. 2.

Inhibition of TFIIS requires both DSIF and NELF and appears to be competitive. (A) Halted U28 complexes were incubated with either DSIF (≈60 nM) or NELF (≈7 nM) or both for 2 min at 30°C before addition of TFIIS (≈50 nM) and removal of samples at 8-s intervals. Samples were separated on a 12.5% denaturing gel. (B) Competitive inhibition of TFIIS by DSIF/NELF. TFIIS-mediated transcript cleavage of C27 TECs was performed as in A with the indicated amounts of TFIIS in the presence of DSIF (≈9 nM) and NELF (≈3.5 nM). The rate of cleavage of C27 TECs (in s^–1) is plotted as a function of TFIIS concentration (•, +DSIF/NELF; ○, –DSIF/NELF). (C) Increasing DSIF/NELF concentration does not increase TFIIS inhibition. Rates of transcript cleavage of C27 TECs were determined as described in the legend to B. TFIIS was at 500 nM. 1× DSIF was ≈9 nM; 1× NELF was ≈3.5 nM.

DSIF/NELF Inhibits TFIIS Binding. In principle, DSIF/NELF could inhibit TFIIS either by interfering with binding of TFIIS to the TEC or by altering the activity of the TEC so that it is less proficient at transcript cleavage. For instance, a subunit of DSIF/NELF could contact the same site of RPB1 that has been found to contribute to TFIIS binding (32), or DSIF/NELF could interact near the active site of RNAPII and prevent TFIIS stabilization of Mg²⁺II binding. To address this question, we examined C27 TECs formed on the template 2 (Fig. 2B). In the absence of DSIF/NELF, TFIIS gave a half-maximal cleavage rate at ≈20 nM, suggesting that TFIIS binds C27 TECs with a K_d of ≈20 nM (Fig. 2B; C27 TECs were present in amounts equivalent to ≈10 pM RNAPII). When DSIF/NELF was present at ≈3.5 nM, inhibition of TFIIS was overcome at high TFIIS concentration (saturation at ≈1 μM TFIIS; apparent K_d ≈275 nM TFIIS; Fig. 2B), as predicted for an effect on TFIIS binding. These results are inconsistent with models in which DSIF/NELF noncompetitively alters TEC activity (because high TFIIS concentration overcomes inhibition) and in which DSIF/NELF blocks TFIIS binding to RNAPII by binding directly to TFIIS (because the inhibitory effects of DSIF/NELF persist even when TFIIS is present in significant excess). Interestingly, the competition between TFIIS and DSIF/NELF is asymmetric: a 6-fold increase in DSIF and NELF concentrations, individually or together, does not increase inhibition of TFIIS (Fig. 2C). This suggests DSIF/NELF either binds to more than one site on RNAPII, only one of which overlaps the TFIIS-binding site, or inhibits TFIIS binding by distorting the TFIIS-binding site (see Discussion).

DSIF/NELF Does Not Inhibit the Intrinsic Cleavage Activity of RNAPII. If DSIF/NELF did inhibit TFIIS-stimulated transcript cleavage by interfering with the catalytic mechanism rather than with TFIIS binding, DSIF/NELF also should inhibit intrinsic transcript cleavage by RNAPII. Although TFIIS stimulates transcript cleavage by stabilizing Mg²⁺II binding, RNAPII can catalyze the same cleavage reaction without TFIIS at slower rates (33). The rate of intrinsic cleavage can be increased by high pH, substitution of Mn²⁺ for Mg²⁺, or high concentrations of Mg²⁺ (26, 33, 34). We reasoned that substitution of Mn²⁺ for Mg²⁺ had the least potential to perturb RNAPII or DSIF/NELF. Therefore, we tested the effect of DSIF/NELF on the rate of Mn²⁺-mediated intrinsic cleavage. DSIF/NELF had no effect on Mn²⁺-mediated intrinsic cleavage when compared side by side with its effect of TFIIS-induced cleavage of the same U28 TECs (Fig. 3; Fig. 6, which is published as supporting information on the PNAS web site). DSIF/NELF also had no effect on the rate of pyrophosphorolysis in U28 TECs (Fig. 3). These results confirm that DSIF/NELF inhibits TFIIS action by interfering with TFIIS binding and not by affecting the catalytic activity of RNAPII.

Fig. 3.

Intrinsic transcript cleavage activity of RNAPII is not inhibited by DSIF/NELF. U28 TECs were subjected to Mn²⁺-induced intrinsic transcript cleavage at 10 mM MnCl₂, TFIIS-induced cleavage (50 nM as a control), or pyrophosphorolysis with 1 mM sodium pyrophosphate (PPi) in the absence or presence of DSIF/NELF. MnCl₂ was replaced with 8 mM MgCl₂ in the TFIIS and PPi reactions. The reaction conditions were otherwise as described in the legend to Fig. 2 A. Some lane-to-lane variation reflects loading differences caused by inhomogeneous sampling of the bead-immobilized TECs. The relative amounts of U28, C27, and A26 RNAs and the rates of U28 cleavage (0.04 ± 0.007 min^–1) are indistinguishable with or without DSIF/NELF (Fig. 6).

Discussion

We report here that DSIF/NELF possesses a previously undescribed activity, inhibition of TFIIS-mediated transcript cleavage, that does not require contact to nascent RNA and thus occurs by a different mechanism than the known inhibitory effect of DSIF/NELF on nucleotide addition. DSIF/NELF inhibition of TFIIS may have important implications for the regulation of early stages of transcript elongation by RNAPII.

DSIF/NELF Inhibits TFIIS Binding Far from Its Proposed Contact with Exiting RNA, Consistent with DSIF/NELF Tethering to RNAPII. Our results establish that DSIF/NELF inhibits TFIIS-stimulated transcript cleavage and strongly suggest that DSIF/NELF inhibits TFIIS binding to RNAPII. Inhibition of binding is likely because (i) the magnitude of TFIIS inhibition is constant even when the basal cleavage rate varies (Fig. 1B); (ii) raising the TFIIS concentration gradually overcomes DSIF/NELF inhibition (Fig. 2B); and (iii) intrinsic transcript cleavage, which occurs by the same catalytic mechanism as TFIIS-stimulated cleavage, is not inhibited by DSIF/NELF (Fig. 3).

Our results do not establish how the six-polypeptide DSIF/NELF complex interferes with TFIIS binding. To consider this question, we compared possible sites of TFIIS competition to the proposed sites of DSIF/NELF inhibition of nucleotide addition and of DSIF binding to RNAPII. DSIF/NELF inhibition of nucleotide addition is thought to be mediated by interactions of RD (NELF-E) with nascent RNA (8, 9, 12). Our finding that DSIF/NELF inhibition of nucleotide addition requires ≥18 nt of nascent RNA (Fig. 1E) is consistent with this view, because both nuclease protection (35) and structural modeling (Fig. 4A and refs. 36 and 37) suggest that 16–18 nucleotides of nascent RNA are protected within RNAPII.

Fig. 4.

Mechanism of DSIF/NELF inhibition of TFIIS. (A) Location of TFIIS binding to a TEC. A yeast RNAPII TEC model based on the crystal structure of Gnatt et al. (37) and the nucleic-acid-scaffold model of Korzheva et al. (36) is depicted with TFIIS located as reported by Kettenberger et al. (23). Portions of RNAPII (Left) are rendered semitransparent to reveal RNA, DNA, and TFIIS in the internal channels of the enzyme. (Left) The approximate length of exiting RNA; (Right) the distance between the exiting RNA and TFIIS. The location at which RPB1 CTD emerges from RNAPII (mobile magenta worm), the proposed SPT5-binding site (blue worm; ref. 39) on the RPB7 subunit (gray worm), the weakly bound Mg²⁺ ion required for transcript cleavage (yellow sphere; Mg²⁺II), and TFIIS (orange worm) are shown in both Left and Right. RPB4 is omitted for clarity. (B) Model depicting how DSIF/NELF inhibition of TFIIS could regulate promoter-proximal pausing. Double-headed arrow depicts possible NELF fluctuation (one of several ways to explain DSIF/NELF competition for TFIIS binding; see text). Red X depicts backtrack-paused RNAPII active site.

In contrast to its RNA length-dependent effect on nucleotide addition, however, DSIF/NELF inhibited TFIIS even in TECs in which the nascent RNA had not yet emerged from RNAPII (e.g., U14 TEC, Fig. 1B). This difference suggests that TFIIS inhibition does not require RD subunit interaction with upstream RNA. Instead, DSIF/NELF appears to interfere with TFIIS binding to RNAPII, which requires that DSIF/NELF obscure or distort the TFIIS interaction site.

The site of TFIIS binding to the TEC is far from the point of nascent RNA exit (Fig. 4A). TFIIS binds on the downstream face of RNAPII, where it inserts a Zn²⁺-ribbon domain into the secondary channel of RNAPII to stabilize binding of the second active-site Mg²⁺ ion required for the transcript cleavage reaction (23, 32). In contrast, the nascent RNA exits on the upstream face, opposite to and >100 Å away from the TFIIS-binding site (Fig. 4A and refs. 36 and 37).

At least three explanations for the widely separated sites of DSIF/NELF action are possible. An elongated DSIF/NELF could span the distance between the exiting RNA and the TFIIS-binding site; DSIF/NELF could bind near the RNA exit channel and allosterically weaken the TFIIS-binding site; or DSIF/NELF could fluctuate between these two locations, tethered to a site between them. All three explanations are consistent with DSIF binding to RPB7 (Fig. 4A), a location proposed because an archaeal SPT4 homologue (Sulfolobus acidocaldarius RpoE) is found fused to RPB7 (4, 38) and because SPT5 may interact with RNAPII through RPB7 (39). However, the failure of 6-fold more DSIF/NELF to increase TFIIS inhibition even though TFIIS can outcompete DSIF/NELF inhibition (Fig. 2 B and C) favors either tethering of DSIF/NELF or a second low-affinity mode of TFIIS binding. Either simple or allosteric competition between TFIIS and DSIF/NELF predicts an increased effect of DSIF/NELF at higher DSIF/NELF concentrations. Even if DSIF/NELF allosterically distorts the TFIIS-binding site, a reciprocal distortion of the DSIF/NELF site should occur upon TFIIS binding. In contrast, if tethered DSIF/NELF obscures TFIIS binding via a secondary weak contact, then the local concentration of DSIF/NELF at the TFIIS-competitive site would be constant, and further increases in DSIF/NELF concentration would have no effect until its bulk concentration approached this local concentration. Thus, tethering of DSIF/NELF to RNAPII, possibly by SPT5 binding to RPB7, can explain both why competition between TFIIS and DSIF/NELF is nonreciprocal and how DSIF/NELF can affect distant parts of a TEC (exiting RNA and TFIIS contact). Alternatively, weak binding of high-concentration TFIIS could reflect a second low-affinity mode of TFIIS binding to TECs that are saturated with nonfluctuating DSIF/NELF at all DSIF/NELF concentrations we tested.

Inhibition of TFIIS by DSIF/NELF May Contribute to Promoter-Proximal Pausing and Negative Regulation of Early Elongation. During the transition from the initial transcription complex to a productive TEC, RNAPII is susceptible to premature termination (40). This susceptibility ends when P-TEFb phosphorylates targets including SPT5, RD (NELF-E), and Ser-2 in the RPB1 CTD, which inactivates NELF or dissociates it from the TEC (13, 14, 41, 42). Luse and coworkers (20, 43) have established that these initial transcription complexes are especially susceptible to prolonged pausing or arrest. Promoter-proximal pausing and susceptibility to arrest persist until the nascent RNAs are longer than ≈50 nt and occur by backtracking of RNAP along the RNA and DNA chains, in part because initial TECs lack nascent RNA structures stable enough to prevent backtracking. Backtracking causes pausing because the RNA 3′ nucleotide is displaced from the active site and leads to arrest when 3′ RNA becomes trapped in the RNAPII secondary channel (also called the pore and funnel; refs. 44 and 45). This extensive promoter-proximal pausing/arrest can be relieved by TFIIS-stimulated cleavage of the backtracked RNA (20).

We suggest that DSIF/NELF inhibition of TFIIS may contribute to control promoter-proximal pausing/arrest and formation of productive TECs (Fig. 4B). Because RNAPII is prone to backtracking in the promoter-proximal region and may depend on TFIIS to avoid becoming trapped in backtracked states, inhibition of TFIIS by DSIF/NELF may increase pausing and arrest simply by preventing TFIIS from rescuing backtracked RNAPII at the promoter-proximal sites (Fig. 4B). Thus, DSIF/NELF may negatively regulate the transition to productive elongation in part by inhibiting TFIIS.

A role of DSIF/NELF inhibition of TFIIS makes sense in vivo. DSIF/NELF directly regulates RNAPII pausing/arrest in the early transcribed regions of Drosophila hsp70 gene and in certain estrogen-responsive human genes (15, 16). Adelman et al. (21) recently reported that both efficient escape of RNAPII from promoter-proximal pauses in hsp70 and heat-shock induction of hsp70 require TFIIS. Thus, DSIF/NELF could contribute to promoter-proximal RNAPII stalling in hsp70 in part by inhibiting TFIIS. This effect may contribute to the regulation of many genes transcribed by RNAPII, for which control of premature termination is a common theme.

However, the complex interplay of regulators that control these steps remains to be elucidated. NELF and the initiation factor TFIIF also inhibit each other (14). TFIIF, which assists promoter escape (46), also inhibits TFIIS-stimulated cleavage (47) but not TFIIS suppression of pausing (48). The points at which NELF can first act on TFIIF or TFIIS and at which NELF is released from the TEC are uncertain. NELF appears to be inactivated by the capping enzyme (19). Thus, DSIF/NELF may delay RNAPII elongation into genes via multiple inhibitory interactions that increase promoter-proximal pausing (inhibition of TFIIF and nucleotide addition) and prevent reactivation of stalled TECs (inhibition of TFIIS) until a regulatory factor (e.g., P-TEFb or Drosophila heat-shock factor), capping enzyme, or both, binds to the TEC, relieves DSIF/NELF inhibition, and allows TFIIS-stimulated transcript cleavage of backtracked RNA in the promoter-proximal, stalled TECs.