Three acts of one performance
By now you have met the players. You know that transcription copies one gene's DNA into RNA, that RNA polymerase is the machine that does the writing, that it reads the template strand and not the coding strand, and that it always grows RNA in the 5'-to-3' direction. This guide stops naming the players and instead watches them work — it follows a single molecule of polymerase through one complete act of copying, from the moment it commits to a gene to the moment it lets go. That whole performance breaks naturally into three acts: initiation (getting started), elongation (the steady middle), and termination (knowing when to stop).
It helps to fix the scale before we begin. The three acts are wildly lopsided in difficulty. Initiation is slow, fussy, and tightly controlled — it can take seconds to minutes of trying, and it is where the cell makes nearly all of its decisions about whether a gene gets used at all. Elongation, by contrast, is fast and businesslike: once a bacterial polymerase is rolling it adds tens of nucleotides every second, copying thousands of bases without pause. Termination is brief but decisive — a clean full stop that defines where the transcript ends. Throughout this guide I will lean on a running picture: a copyist who must find the right page in a huge book, settle into a writing rhythm, and then know to put down the pen.
Act one — initiation: finding the start and stuttering into motion
Initiation begins with a search. A genome is millions of base pairs long and chemically rather uniform, so the polymerase must locate the handful that mark a gene's start. In bacteria the complete enzyme (the core plus a swappable sigma factor) slides and bumps along the DNA until the sigma factor recognizes a promoter — the road sign, sitting just upstream, that says 'start here, on this strand, in this direction.' You met the promoter in the previous guide; the key new idea is that finding it is the rate-limiting hurdle of the whole process. Everything downstream is fast; this first step is where the polymerase spends its time.
Once the polymerase has docked on the promoter, two things must happen before it can write a single letter. First it sits down on the closed double helix — a state called the *closed complex*, where the DNA is still fully zipped. Then it pries the two strands apart over a short stretch, about a dozen base pairs, exposing the template — the *open complex*, also called the transcription bubble. Only now can the catalytic site reach the template bases. The enzyme lines up the first two incoming ribonucleotides opposite the template, joins them, and begins to build RNA. By convention the very first base copied is numbered +1, the transcription start site; bases before it (toward the promoter) are 'upstream' with minus numbers, bases after it 'downstream'.
Here comes the surprise that beginners rarely expect: starting is not a clean one-shot event. While the polymerase is still clamped to the promoter, it tends to *stutter* — it makes a short RNA of a few nucleotides, lets it slip out, and tries again, over and over, like a writer crumpling up false starts. This wasteful sputtering is called [[abortive-initiation-and-promoter-clearance|abortive initiation]]. The enzyme is reluctant to release its tight grip on the promoter, so it scrunches the downstream DNA inward to keep making RNA without moving — and most of those tiny transcripts fall apart. Only when one transcript grows long enough (roughly 8 to 12 nucleotides) does the enzyme finally break its hold on the sigma factor and the promoter, an event called promoter escape or *promoter clearance*. That escape is the true threshold between merely trying and actually transcribing.
Act two — elongation: read a base, add a base, slide along
With the promoter behind it, the polymerase relaxes into the steady rhythm of elongation. In bacteria the sigma factor now drops off (it was only needed to find the start), leaving the leaner core enzyme to do the long-distance copying. Picture the enzyme as a moving bubble gliding along the gene: ahead of it the double helix is unwound to expose fresh template, and behind it the two DNA strands re-zip and the finished RNA peels away. The bubble does not grow — it *travels*, opening new DNA in front at exactly the rate it closes old DNA behind, a self-contained machine that carries its own little patch of melted DNA along for the ride.
Inside the bubble, the chemistry repeats like clockwork. At each step the enzyme reads the next template base, lets a matching ribonucleoside triphosphate (ATP, GTP, CTP, or UTP) test-fit against it — A pairs with the template's T as U, G pairs with C, and so on — and if it fits, forges a phosphodiester bond onto the growing RNA's 3' end, releasing two phosphates as the energy source. Then the whole enzyme ratchets forward exactly one base and repeats. A short stretch of the newest RNA, about 8 or 9 nucleotides, stays paired to the template as an RNA-DNA hybrid before peeling off; the rest of the transcript trails out behind. And it is quick — a bacterial polymerase manages on the order of tens of nucleotides per second.
moving ---> bubble travels along the gene
rewound DNA | transcription bubble | DNA to be read
============= ( unwound, ~13 bp open ) =================
3'...A T G C [ T A C G G A T ] G C A...5' <- template (read 3'->5')
| | | | | |
5'...U A U G C C U-OH (3' growing end) <- new RNA
============= =================
(re-zipped) RNA peels off here (still paired)Elongation is not flawless, and it does not pretend to be. The polymerase carries only *modest* proofreading: if it mistakenly adds a wrong base, it can pause, back up a step, and clip off the offending end before trying again. That trims the error rate to roughly one mistake in 10,000 to 100,000 bases — noticeably sloppier than DNA replication's near-perfection. And that looseness is perfectly acceptable, for an honest reason worth holding onto: an RNA copy is disposable and made in many copies, so one flawed transcript among thousands is no catastrophe, whereas a replication error would be inherited by every descendant cell. The cell spends its accuracy budget where mistakes are permanent.
Act three — termination: two ways for bacteria to stop
Knowing when to stop matters as much as knowing when to start. Without a reliable full stop, the polymerase would barrel onward into the next gene, and the next, producing one giant tangled transcript that is good for nothing. Termination is the cell's period at the end of the sentence: the signal and the mechanism that make the polymerase release its RNA, let go of the DNA, and quit. Bacteria solve this two different ways, and both are worth seeing because they show the same goal reached by completely different tricks.
The first is intrinsic termination, also called factor-independent because it needs no extra protein — the stop signal is written into the RNA itself. As the polymerase transcribes a particular stretch, the freshly made RNA contains a self-complementary, GC-rich sequence that immediately folds back on itself into a tight hairpin (a stem-loop). Right after the hairpin comes a run of about six or more uracils, so the RNA's 3' end is held to the template only by weak rU-dA base pairs. The combination is fatal to the grip: the hairpin tugs at and destabilizes the polymerase just as the feeble U-A pairs let the RNA slip free. The enzyme stalls, the RNA peels off, transcription ends. No outside help required — the transcript carries its own stop sign.
The second is [[rho-dependent-termination|Rho-dependent termination]], which does need a helper: a ring-shaped protein called Rho. Rho latches onto the growing RNA at a specific landing patch and then chases the polymerase down the transcript, using ATP to pull itself along the RNA — picture a runner sprinting to catch a train. The polymerase, meanwhile, runs ahead but periodically pauses (often at a terminator-like sequence). When Rho catches up to a stalled polymerase, it pries the RNA-DNA hybrid apart and forces the transcript loose. Same outcome as the hairpin route — released RNA, freed enzyme — but achieved by an active protein hunting the polymerase down rather than by a fold in the RNA.
Stepping back — the shape of the whole process
- Search and bind: the polymerase (with its sigma factor in bacteria) scans the DNA and recognizes a promoter — the slow, rate-limiting, heavily regulated step.
- Open and start: it melts a small bubble, exposes the template, and joins the first nucleotides — stuttering through abortive starts until it finally escapes the promoter.
- Roll along: now in elongation, the bubble travels, reading 3'->5' and writing RNA 5'->3' at tens of bases per second, re-zipping DNA behind and proofreading modestly.
- Stop and release: at the end it terminates — by an intrinsic hairpin-plus-uracils signal, or by the Rho protein catching the enzyme — letting the finished transcript go.
Two honest pointers before you climb on. First, eukaryotes do not work quite like this. Their initiation is far more involved — a crowd of general transcription factors and a pre-initiation complex must assemble before RNA polymerase II can begin — and their termination is comparatively *sloppy*: Pol II has no neat hairpin stop sign and often runs hundreds of bases past the real end of the message, with termination tied to cutting and tailing the RNA rather than to a crisp signal. The clean three-act story above is the bacterial version; treat it as the clearest example, not a universal blueprint.
Second, do not let the tidy three-act framing fool you into thinking the acts are equal or independent. Initiation is overwhelmingly where control lives — activators, repressors, sigma factors, and (in eukaryotes) transcription factors all pile their effort onto the decision of whether starting happens at all, which is why this stage is the cell's main switch for gene expression. Elongation and termination are mostly the machine running its course, though even they offer regulatory openings (a polymerase can be made to pause, or termination can be triggered early to cut a transcript short). With the three stages in hand, you are ready for the next rung's question: what happens to the raw RNA *after* the polymerase lets it go.