The Polymerase Chain Reaction (PCR)

A single strand of DNA can become a billion copies in an afternoon. That is the promise Kary Mullis reportedly worked out while driving down a moonlit California highway in 1983. It is one of the few moments in biochemistry's history that can be pinned to a single conceptual leap. Most breakthroughs instead come from decades of incremental refinement. The idea is disarmingly simple. Use two short oligonucleotide primers to bracket a region of DNA. Use a DNA polymerase to copy it. Then repeat the copying step over and over. Each round doubles the number of target molecules. Because the doubling compounds exponentially, twenty or thirty cycles change things dramatically. An undetectable trace of DNA becomes a macroscopic, sequenceable, clonable quantity. The reason PCR matters is not merely that it amplifies DNA. Cells already do that every time they divide. The reason is that PCR lets you choose exactly which sequence to amplify. You can pick one target out of a genome of three billion base pairs. This works using nothing but the sequence information in two primers. Each primer is only 18–25 nucleotides long. This turns Hybridization and the Basis of Nucleic Acid Detection from a passive detection tool into an active engineering tool. Instead of asking "is this sequence present," PCR lets you manufacture as much of that sequence as you want. Diagnostics, forensics, ancient-DNA paleogenomics, cloning, mutagenesis, and the entire modern genomics pipeline all rest on this single reaction. The big picture is a thermodynamic and enzymatic loop repeated in a tube. Heat separates the double helix into single strands. This is just controlled application of the physics of Denaturation and Renaturation of DNA. Cooling allows short synthetic primers to find and hybridize to their complementary sequences. This exploits the same base-pairing thermodynamics that governs any nucleic acid hybridization. A thermostable DNA polymerase then extends each primer. The most famous example is Taq polymerase, from the hot-spring bacterium Thermus aquaticus. It uses the four deoxynucleoside triphosphates (dNTPs) as substrates. Repeating this three-temperature cycle — melt, anneal, extend — converts one double-stranded template into two, then four, then eight. The amount doubles geometrically until reagents become limiting. This article works through the mechanism of primer extension at the polymerase active site. It covers the exponential kinetics of amplification and why they eventually saturate. It also explains the biophysics of primer design and annealing temperature. It reviews the practical chemistry (Mg²⁺, dNTPs, buffer, thermostability) that makes 30 cycles of 95 °C survivable for an enzyme. It closes with the worked arithmetic every PCR user needs. This includes cycle number versus yield, primer $T_m$ calculations, and the logic of quantitative PCR. Finally, it points toward how these same amplified fragments feed into DNA Sequencing: Sanger and Next-Generation Methods.

Key takeaways

PCR amplifies a chosen DNA segment using two primers that flank it and a thermostable polymerase that extends them.
Each cycle has three thermal steps: denaturation (~94–98 °C), annealing (~50–65 °C), extension (~72 °C).
Amplification is exponential, $N = N_0(1+E)^n$ , until reagents or polymerase processivity become limiting (plateau phase).
Taq polymerase survives repeated 95 °C denaturation because its hyperthermophilic origin gives it a rigid, disulfide- and salt-bridge-stabilized fold; it lacks 3′→5′ proofreading, giving ~1 error per 10⁴–10⁵ bases.
Primer annealing temperature is set below the primer $T_m$ , calculated from GC content and salt concentration by empirical formulas (Wallace rule, nearest-neighbor thermodynamics).
Quantitative PCR (qPCR) turns the exponential phase into a measurement of initial template copy number via the cycle threshold $C_t$ .
Key idea: exponential doubling driven by primer-directed, temperature-cycled DNA synthesis converts a trace of target sequence into a macroscopic, sequence-defined population of copies.

The core insight

Two short primers define the boundaries of amplification; a thermostable polymerase copies between them; repeated melt–anneal–extend cycles double the target DNA exponentially, $N=N_0 2^n$ in the ideal limit, until substrate depletion forces a plateau.

Where this lives today

PCR is now the substrate-agnostic workhorse of molecular biology and clinical diagnostics. Real-time quantitative PCR (qPCR), using fluorescent intercalating dyes (SYBR Green) or sequence-specific hydrolysis probes (TaqMan probes, exploiting the 5′→3′ exonuclease activity of Taq polymerase to cleave a quenched fluorescent probe during extension), turns the exponential phase itself into a quantitative assay for viral load, gene expression level, or pathogen copy number — this is precisely the calculation carried out in Worked Example 3 above, performed routinely in clinical virology labs (including for SARS-CoV-2 detection during the COVID-19 pandemic, where $C_t$ values were reported directly to clinicians as a proxy for viral load). Reverse-transcription PCR (RT-PCR) couples a reverse transcriptase step ahead of thermal cycling to amplify RNA targets by first converting them to complementary DNA. Multiplex PCR uses several primer pairs simultaneously in one tube to genotype multiple loci at once, exploited in forensic short tandem repeat (STR) profiling, where amplified length polymorphisms at a panel of loci generate the numeric genetic profiles used in criminal investigation and paternity testing. Digital PCR partitions a sample into thousands of nanoliter-scale droplets or wells so that each partition contains zero or one template molecule, converting the analog exponential-amplification measurement into a direct molecular counting experiment (Poisson statistics on the fraction of positive partitions), giving absolute copy-number quantification without needing a standard curve at all. Site-directed mutagenesis protocols use PCR with mismatched primers to introduce deliberate point mutations, insertions, or deletions into cloned genes. Ancient-DNA and forensic laboratories rely on PCR's exquisite sensitivity — in principle amplifiable from a single template molecule — to recover genetic information from highly degraded samples, a use case that pushed the field toward hot-start polymerases and careful primer design to suppress the non-specific amplification that becomes especially problematic at very low input copy number. Downstream, the DNA fragments PCR produces are the direct feedstock for DNA Sequencing: Sanger and Next-Generation Methods, where amplified, purified, sequence-defined DNA populations are exactly what both Sanger dideoxy sequencing and next-generation sequencing platforms require as starting material.

Ask the tutor: How does adding a fluorescent TaqMan probe with a quencher let qPCR generate a real-time signal exactly proportional to the amount of new DNA synthesized each cycle?

The physical picture — thorough prose

A PCR reaction tube looks unremarkable: a colorless solution containing template DNA (perhaps a single copy hidden among a microgram of human genomic DNA), two synthetic oligonucleotide primers present in vast molar excess over the template, all four dNTPs, Mg²⁺ ions, a reaction buffer near pH 8.3, and a thermostable DNA polymerase. What makes it remarkable is the thermal program imposed on it: a machine (a thermal cycler) drives the tube through 25–35 repetitions of a three-temperature cycle, each lasting on the order of a minute. In the first step, denaturation, the tube is heated to 94–98 °C. This is far above the melting temperature $T_m$ of genomic duplex DNA (~85–95 °C depending on GC content and salt), so hydrogen bonds between base pairs and the stacking interactions between adjacent bases are overcome, exactly as in the thermal denaturation of any duplex. The double helix separates into two single strands. In the second step, annealing, the temperature drops to 45–65 °C. At this temperature, the long genomic strands are too complex to efficiently re-anneal to each other within the short annealing time (their effective concentration of any specific complementary segment is minuscule), but the short synthetic primers — present at ~1000-fold molar excess over any genomic sequence — rapidly find and hybridize to their complementary sites, following the same nucleation-and-zippering kinetics that govern any short-oligonucleotide hybridization. In the third step, extension, the temperature rises to about 72 °C, the temperature optimum for Taq polymerase, and the enzyme extends each annealed primer 5′→3′ using the template strand as a guide and the four dNTPs as substrates, releasing pyrophosphate at every step. The critical geometric trick is that the two primers are oriented so their 3′ ends point toward each other, each on a different template strand, so that after the first cycle each new strand has a defined 3′ end exactly at the position where the other primer will bind. From the second cycle onward, newly synthesized strands themselves become templates whose ends are precisely bounded by the two primer sites, so only the sequence between and including the primers is amplified — everything outside that window is diluted out exponentially relative to the target because it is copied only linearly (from the original template) rather than exponentially.

Figure

Three-step thermal cycle and primer geometry

Once new strands with defined ends accumulate — after roughly two to three cycles — the population is dominated by these short "amplicon" molecules bounded exactly by the primer sequences, and the reaction settles into clean exponential doubling for many cycles, until dNTP depletion, primer exhaustion, polymerase inactivation, or product reannealing (which competes with primer annealing at high product concentration) cause the reaction to plateau. The polymerase itself works by the universal two-metal-ion mechanism shared by all template-dependent nucleic acid polymerases. Two Mg²⁺ ions in the active site coordinate the 3′-OH of the primer terminus, the α-phosphate of the incoming dNTP, and key active-site carboxylates (conserved Asp residues in the palm domain). Metal A activates the primer 3′-OH for in-line nucleophilic attack on the α-phosphorus; metal B stabilizes the developing negative charge on the leaving pyrophosphate group and orients the triphosphate. The polymerase only proceeds with high fidelity if the incoming dNTP is correctly Watson–Crick paired with the templating base, because the fingers domain closes around a correctly shaped base pair — an induced-fit checkpoint that discriminates against mismatches by roughly three to four orders of magnitude in rate.

Figure

Two-metal-ion mechanism at the polymerase active site

The core ideas — governing equations

The central quantitative statement of PCR is the amplification law:

N_n = N_0 (1+E)^n

where:

$N_n$ = number of target amplicon molecules after $n$ cycles
$N_0$ = number of target molecules at the start (the quantity you often want to infer)
$E$ = per-cycle amplification efficiency, $0 \le E \le 1$ ( $E=1$ is perfect doubling)
$n$ = number of thermal cycles completed

Reading: every cycle multiplies the amount of specific product by a factor $(1+E)$ ; with ideal efficiency ( $E=1$ ) this is simple doubling, $N_n = N_0\,2^n$ . Real reactions have $E$ slightly below 1 because of imperfect primer annealing, polymerase processivity limits, and (in later cycles) substrate depletion. The annealing temperature is chosen relative to primer melting temperature, itself estimated by empirical rules. The simplest, the Wallace rule, for short primers (<14 nt, though widely misused for longer ones):

T_m \approx 4(G+C) + 2(A+T)\ \ (^\circ\text{C})

where:

$G, C, A, T$ = counts of each base in the primer sequence
the coefficients reflect the extra hydrogen bond (3 vs 2) and stacking energy of G·C pairs relative to A·T pairs

For longer primers a nearest-neighbor thermodynamic model is used instead:

T_m = \frac{\Delta H^\circ}{\Delta S^\circ + R\ln(C_T/4)} - 273.15

where:

$\Delta H^\circ, \Delta S^\circ$ = summed nearest-neighbor enthalpy and entropy parameters for the primer sequence (empirically tabulated per dinucleotide step)
$R$ = gas constant, 1.987 cal·mol⁻¹·K⁻¹
$C_T$ = total strand concentration (primer concentration, since template is negligible), with the $/4$ factor applying for non-self-complementary duplexes at equal strand concentrations
the result converts from kelvin to Celsius

Reading: this is the same van 't Hoff-type melting expression used for any two-state duplex; PCR primer design simply applies it to short synthetic oligomers, then chooses an annealing temperature about 3–5 °C below the lower of the two primers' $T_m$ to ensure specific, efficient hybridization without excessive off-target binding.

Derivation AThe exponential amplification law from first principles

Goal. Show that if each cycle converts a fraction $E$ of single-stranded template into extended, primer-bound duplex product, the target sequence count obeys $N_n = N_0(1+E)^n$ , and identify why $E$ decays in late cycles (the plateau).

Step 1 — Define what "one cycle" does to the amplicon population. Let $N_n$ be the number of double-stranded amplicon molecules (each strand bounded by a primer at each end) present just before cycle $n+1$ begins. During denaturation, each duplex separates into two single strands, giving $2N_n$ single strands available as templates. During annealing and extension, each single strand can be converted into a new duplex-forming strand if (i) a primer successfully anneals to it and (ii) the polymerase successfully extends that primer to completion before the cycle ends. Let $E$ be the combined probability per strand per cycle that both of these succeed. Then the number of strands successfully extended in this cycle is:

\Delta N = E \times (2N_n)/2 = E\,N_n

Here the factor of two from denaturation and the factor of two from having two complementary strands both being extendable cancel in a symmetric idealization (both primers behave identically), leaving a clean fractional increase per existing duplex-equivalent.

Step 2 — Write the recursion. The new duplex count is the old count plus this increment:

N_{n+1} = N_n + E\,N_n = N_n(1+E)

Step 3 — Solve the recursion by iteration. Applying Step 2 repeatedly from $N_0$ :

N_1 = N_0(1+E), \quad N_2 = N_1(1+E) = N_0(1+E)^2, \quad \ldots

N_n = N_0(1+E)^n

This is the amplification law. When $E=1$ (every strand perfectly primed and fully extended), $N_n = N_0\,2^n$ : an ideal doubling every cycle.

Step 4 — Why $E$ falls below 1, and why it eventually collapses toward 0 (the plateau). Early in the reaction, primers and dNTPs are in such vast molar excess over product that annealing is fast and extension always completes within the allotted time — $E$ is close to 1 (typically 0.7–0.95 in practice, limited by annealing/extension kinetics and enzyme processivity). As cycles proceed, three things erode $E$ : 1. Substrate depletion. dNTPs and primers are consumed stoichiometrically; once product concentration approaches primer or dNTP concentration (typically after product reaches ~10⁻⁸–10⁻⁷ M, i.e., roughly $10^{12}$ – $10^{13}$ molecules in a 25–50 µL reaction), the bimolecular annealing step slows because primer is no longer in overwhelming excess. 2. Product reannealing competition. At high copy number, complementary product strands can find each other and reanneal directly (a bimolecular process whose rate scales with product concentration squared), out-competing primer annealing, which scales only linearly with primer concentration. This is the same competitive hybridization kinetics that governs any renaturation reaction — the strand most concentrated wins the kinetic race. 3. Enzyme inactivation and processivity. Thermostable polymerase slowly loses activity from repeated exposure to 94–98 °C, and long templates are not always fully extended within the extension time window, so a growing fraction of "attempts" fail to produce full-length product. As any of these effects grow, $E \to 0$ and $N_{n+1} \approx N_n$ : the reaction plateaus regardless of how many further cycles are run. This is why PCR yield is not simply "more cycles equals more product" beyond roughly 30–35 cycles — the useful, quantitatively informative regime is the early exponential phase, which is exactly the phase exploited by quantitative PCR (qPCR).

Go deeper · Derivation B · 2 minMelting temperature of a short primer from nearest-neighbor thermodynamics, and the annealing-temperature design rule

Derivation BMelting temperature of a short primer from nearest-neighbor thermodynamics, and the annealing-temperature design rule

Goal. Derive the two-state van 't Hoff relation for primer $T_m$ used in step 2 above, and show how it sets the practical annealing temperature.

Step 1 — Treat primer–template hybridization as a two-state equilibrium. For a short primer P annealing to its complementary template strand T, assume only fully duplex or fully single-stranded states are populated (a good approximation for primers under ~30 nt, which show sharp, cooperative, "all-or-none" melting because there is no stable partially-formed intermediate structure):

\text{P} + \text{T} \rightleftharpoons \text{PT}, \qquad K_{eq} = \frac{[\text{PT}]}{[\text{P}][\text{T}]}

Step 2 — Express $K_{eq}$ via standard thermodynamics. Since $\Delta G^\circ = -RT\ln K_{eq} = \Delta H^\circ - T\Delta S^\circ$ , rearranging gives:

\ln K_{eq} = -\frac{\Delta H^\circ}{RT} + \frac{\Delta S^\circ}{R}

Step 3 — Impose the melting condition. Define $T_m$ as the temperature at which exactly half of the strands are in duplex form. For non-self-complementary strands present at equal total concentration $C_T$ (with primer in large excess over any specific template site, primer concentration effectively sets $C_T$ ), the half-dissociated condition gives:

K_{eq}(T_m) = \frac{4}{C_T}

(this factor of 4 arises from the stoichiometry of two distinct, equal-concentration single strands reaching 50% pairing; a self-complementary duplex instead gives a factor of 1).

Step 4 — Substitute and solve for $T_m$ . Insert $K_{eq}(T_m) = 4/C_T$ into the van 't Hoff expression from Step 2:

\ln(4/C_T) = -\frac{\Delta H^\circ}{RT_m} + \frac{\Delta S^\circ}{R}

Rearranging for $T_m$ :

T_m = \frac{\Delta H^\circ}{\Delta S^\circ - R\ln(4/C_T)} = \frac{\Delta H^\circ}{\Delta S^\circ + R\ln(C_T/4)}

which is exactly the nearest-neighbor formula stated above (converted to Celsius by subtracting 273.15). $\Delta H^\circ$ and $\Delta S^\circ$ are obtained by summing tabulated nearest-neighbor stacking parameters (e.g., for each AA, AT, TA, CA, GT, CT, GA, CG, GC, GG dinucleotide step) across the primer sequence, plus helix-initiation terms.

Step 5 — From $T_m$ to a practical annealing temperature. Both primers are designed so their nearest-neighbor $T_m$ values are close to each other (within ~2–3 °C) and typically fall in the 55–65 °C range for GC content near 50%. The annealing temperature $T_a$ used in the thermal cycler is chosen a few degrees below the lower of the two $T_m$ values:

T_a \approx T_m(\text{lower primer}) - (3\text{–}5\,^\circ\text{C})

Choosing $T_a$ too high loses specific product (primers cannot stay bound long enough for extension to initiate); choosing $T_a$ too low permits mismatched, off-target annealing, producing spurious bands. This is a direct, practical corollary of the same duplex-stability physics that governs Hybridization and the Basis of Nucleic Acid Detection: stringency is controlled by how far the working temperature sits below the calculated $T_m$ .

Worked examples

Example 1 — Foundational: exponential yield after a defined number of cycles. Suppose a PCR reaction starts with exactly $N_0 = 5$ copies of target DNA (a realistic low-input forensic or ancient-DNA scenario) and the empirically measured per-cycle efficiency, from a standard curve, is $E = 0.85$ (not the idealized $E=1$ , deliberately asymmetric to test the formula honestly). How many target molecules are present after $n=28$ cycles, assuming the exponential regime holds throughout (a reasonable assumption since with only 5 starting copies the reaction is nowhere near saturation at cycle 28)? Apply the amplification law:

N_{28} = N_0(1+E)^{28} = 5 \times (1.85)^{28}

Compute $(1.85)^{28}$ by successive squaring: $1.85^2 = 3.4225$ ; $1.85^4 = 3.4225^2 = 11.7135$ ; $1.85^8 = 11.7135^2 \approx 137.21$ ; $1.85^{16} \approx 137.21^2 \approx 18,826$ ; $1.85^{24} = 1.85^{16}\times 1.85^8 \approx 18{,}826 \times 137.21 \approx 2.583\times10^{6}$ ; $1.85^{28} = 1.85^{24}\times1.85^4 \approx 2.583\times10^6 \times 11.7135 \approx 3.026\times10^{7}$ .

N_{28} \approx 5 \times 3.026\times10^{7} \approx 1.51\times10^{8}\ \text{molecules}

Limiting-case check: if instead $E=1$ (ideal doubling), $N_{28} = 5\times 2^{28} = 5\times 2.68\times10^8 \approx 1.34\times10^9$ — about 9-fold higher than the $E=0.85$ result, which makes sense: the imperfect efficiency compounds a shortfall every cycle, and $(1.85/2)^{28} \approx 0.112$ , close to the ratio $1.51\times10^8/1.34\times10^9 \approx 0.113$ obtained above, confirming internal consistency of the exponential formula.

Example 2 — Applied: designing a primer pair and choosing the annealing temperature. A researcher wants to amplify a 420 bp fragment of a bacterial gene and designs a forward primer 5′-GCTACGGTTACCGATGCTAA-3′ (20 nt) and a reverse primer 5′-TTGGCATCGTAGCCTTAGCA-3′ (20 nt). Using the Wallace rule for a first estimate: Forward primer base count: G=5, C=4, A=6, T=5 (verify: 5+4+6+5=20 ✓). Wallace estimate:

T_m^{F} = 4(G+C) + 2(A+T) = 4(9) + 2(11) = 36+22 = 58\,^\circ\text{C}

Reverse primer base count: T=4, G=5, C=5, A=6 (4+5+5+6=20 ✓; note this reverse primer happens to have the same GC count coincidentally close to the forward one — deliberately checking with distinct sequences rather than perfectly complementary reverse-complement pairs, since real primer pairs are rarely symmetric):

T_m^{R} = 4(10) + 2(10) = 40+20 = 60\,^\circ\text{C}

The lower of the two estimated $T_m$ values is 58 °C (forward primer). A nearest-neighbor calculation would typically refine this by a few degrees (Wallace systematically overestimates for primers longer than ~14 nt because it ignores sequence-dependent stacking and salt effects), but taking the Wallace estimate as a starting point, the annealing temperature is chosen:

T_a \approx 58 - 4 = 54\,^\circ\text{C}

The cycler program would then run: 95 °C for 30 s (denature), 54 °C for 30 s (anneal), 72 °C for ~21 s (extension, using the common rule of thumb of roughly 1 kb per minute for Taq polymerase, so 420 bp needs well under a minute — in practice rounded up to 30 s for reliability), repeated for 30–35 cycles.

Limiting-case check: if both primers had identical Wallace $T_m$ (a symmetric special case), the rule reduces trivially to $T_a = T_m - 4$ with no ambiguity about "which" $T_m$ to use — confirming that the general rule (always use the lower value) correctly subsumes the equal- $T_m$ case as $T_m^F \to T_m^R$ .

Example 3 — Exam/competition level: back-calculating initial copy number from a qPCR cycle threshold, with an efficiency correction. In a quantitative PCR (qPCR) experiment, a fluorescent reporter's signal crosses a fixed detection threshold at cycle $C_t = 22.7$ for an unknown sample, and a reference standard of exactly $10^6$ copies gives $C_t = 15.3$ under the same amplification efficiency $E=0.92$ (measured independently from a dilution series' standard curve, so it is not assumed to be 1). Find the initial copy number $N_0$ of the unknown. At the threshold cycle, all reactions have accumulated the same absolute number of amplicon molecules $N_{thresh}$ (that is the operational definition of a fixed fluorescence threshold), so:

N_{thresh} = N_0^{unknown}(1+E)^{C_t^{unknown}} = N_0^{ref}(1+E)^{C_t^{ref}}

Solve for the unknown initial copy number:

N_0^{unknown} = N_0^{ref}\,(1+E)^{\,C_t^{ref}-C_t^{unknown}}

Substitute values: $C_t^{ref} - C_t^{unknown} = 15.3 - 22.7 = -7.4$ , and $1+E = 1.92$ .

N_0^{unknown} = 10^6 \times (1.92)^{-7.4}

Compute $(1.92)^{7.4}$ : $\ln(1.92) \approx 0.6523$ ; $7.4 \times 0.6523 \approx 4.827$ ; $e^{4.827} \approx 124.7$ .

N_0^{unknown} = 10^6 / 124.7 \approx 8.02\times10^{3}\ \text{copies}

So the unknown sample contained approximately 8,020 copies of target DNA at the start of the reaction.

Limiting-case check: if the efficiency were ideal ( $E=1$ , so $1+E=2$ ), the same $\Delta C_t = -7.4$ would give $N_0^{unknown} = 10^6\times 2^{-7.4} = 10^6/168.9 \approx 5{,}920$ copies — noticeably different from the $E=0.92$ -corrected value of 8,020. This 26% discrepancy is exactly why real qPCR quantification always measures and applies the actual per-assay efficiency from a standard curve rather than assuming perfect doubling; ignoring a modest efficiency deficit systematically underestimates initial copy number.

Limiting cases, scope & common traps

Regime / question	Behavior	Why
Very early cycles (n = 1–3)	Not yet clean exponential; mixture of original long templates and newly bounded amplicons	Defined-length product only dominates once both primer-bounded ends have been "set," which takes ~2–3 cycles
Mid cycles, ample reagents	Clean exponential, $E \approx 0.8$ – $0.95$	Primers/dNTPs in vast excess; annealing and extension both fast and reliable
Late cycles (typically >30)	Plateau, $E\to 0$	dNTP/primer depletion, product reannealing competes with primer annealing, enzyme activity decays
$E \to 1$ limit	$N_n \to N_0 2^n$	Idealized perfect doubling; useful ceiling, rarely fully achieved
Annealing temperature too high	Weak/no product	Primers cannot remain bound long enough to be extended
Annealing temperature too low	Multiple non-specific bands	Mismatched hybridization tolerated; stringency lost
Extension time too short for target length	Truncated products, low yield	Polymerase (~30–100 nt/s for Taq) cannot reach the far primer before cooling
High-GC template regions	Incomplete denaturation, "GC bubbles" persist	Requires denaturation additives (DMSO, betaine) or higher denaturation temperature
Excess cycles beyond plateau	No further gain, sometimes signal decline	Enzyme degradation, non-specific product competition

The chemist behind it

Kary Banks Mullis conceived PCR in 1983 while working at Cetus Corporation, a Bay Area biotechnology company, where he was tasked with synthesizing oligonucleotides for other scientists' hybridization and sequencing projects — a role that gave him deep, hands-on familiarity with primer annealing chemistry just before he needed it. His insight was not a new enzyme or a new chemical reaction; every individual step (denaturation, primer hybridization, polymerase extension) was already textbook biochemistry. The innovation was the iterative, exponential application of these steps using two primers oriented to bracket a target, replacing what had been an essentially linear, one-round primer-extension technique (used earlier for genetic analysis by others, including Kjell Kleppe in the 1970s, who described the concept without exploiting repeated cycling for exponential gain) with a geometrically compounding amplification scheme. Mullis initially used the Klenow fragment of E. coli DNA polymerase I, which is heat-labile and had to be replenished by hand after every single heat-denaturation step — an exhausting, error-prone, and unscalable procedure. The transformative second step, contributed by others at Cetus (notably Randall Saiki, Henry Erlich, and colleagues, working with polymerase supplied through collaboration involving Thermus aquaticus research), was to substitute a thermostable polymerase purified from the hot-spring bacterium Thermus aquaticus (hence "Taq polymerase"), which survives repeated cycles at 94–98 °C and needs adding only once. This single substitution converted PCR from a laborious manual technique into a fully automatable one, enabling the thermal cycler as an instrument category and PCR as a routine laboratory operation performed millions of times daily worldwide.

Origins · 1 minKary Banks Mullis and how this was discovered

Kary Banks Mullis

1944–2019 · Cetus Corporation, Emeryville, California, early 1980s

A biochemist rather than a physicist by training, Mullis received the 1993 Nobel Prize in Chemistry (shared with Michael Smith, honored separately for site-directed mutagenesis) for the invention of PCR. His path to the idea combined an unusually broad chemical intuition — gained partly through his role synthesizing DNA oligonucleotides at Cetus — with a willingness to ask whether an old two-primer, single-round extension trick used for point mutation detection could be repeated exponentially rather than performed once.

History

- 1971 — Kjell Kleppe and H. Gobind Khorana describe primer-directed DNA repair synthesis in principle, without exploiting repeated cycling.

- 1983 — Kary Mullis conceives the exponential, cyclic primer-extension scheme at Cetus Corporation.

- 1985 — First published PCR paper (Saiki, Scharf, Faloona, Mullis et al.) demonstrates amplification and use in prenatal sickle-cell diagnosis, using Klenow fragment with manual reagent replenishment each cycle.

- 1988 — Thermostable Taq polymerase from Thermus aquaticus is introduced into PCR, enabling automated thermal cycling (Saiki et al., Science).

- 1993 — Mullis awarded the Nobel Prize in Chemistry for the invention of PCR.

Why does PCR need two primers instead of one?

Two primers, oriented so their 3′ ends point toward each other on opposite template strands, bracket the target region and allow synthesis to proceed inward from both sides. This is what makes amplification exponential rather than linear: each new strand made in one cycle becomes, in the next cycle, a template precisely bounded by the position of the other primer, so it too gets copied. A single primer would only ever copy from the original template linearly (one new strand per template per cycle), never compounding, and would also fail to define a fixed downstream boundary for the product.

What actually goes wrong if the annealing temperature is set too low?

Primers can hybridize even to sequences that are not perfectly complementary, because at low temperature the free-energy penalty for a mismatch is more easily offset by the overall binding enthalpy of the rest of the primer. This produces non-specific annealing at multiple genomic sites, which the polymerase will happily extend, yielding extra, unwanted bands on a gel (or spurious fluorescence signal in qPCR). The fix is either to raise the annealing temperature closer to (but still below) the calculated primer $T_m$ , or to use a "hot-start" polymerase that remains inactive until the first high-temperature denaturation step, preventing premature low-stringency extension before cycling even begins.

Why is Mg²⁺ concentration so critical to PCR success?

Mg²⁺ is an essential cofactor for the two-metal-ion nucleotidyl transfer mechanism used by all DNA polymerases: one Mg²⁺ activates the primer's 3′-OH nucleophile and the other stabilizes the developing negative charge on the pyrophosphate leaving group. Mg²⁺ also affects primer-template duplex stability (shielding phosphate backbone charge repulsion) and polymerase-DNA affinity. Too little Mg²⁺ starves the active site and yields little product; too much Mg²⁺ stabilizes mismatched primer binding and reduces polymerase fidelity, increasing non-specific amplification — so PCR protocols typically titrate Mg²⁺ concentration (commonly 1.5–3 mM) for each new primer pair.

How is qPCR different from ordinary endpoint PCR?

Ordinary PCR is read out at the end, after the reaction has already reached its plateau, so the final band intensity mostly reflects the plateau chemistry rather than the original template amount — it is good for detecting presence/absence but poor for quantification. qPCR instead monitors fluorescence every cycle, so it captures the reaction while still in its clean exponential phase, and uses the cycle number at which a fixed fluorescence threshold is crossed ( $C_t$ ) to back-calculate initial copy number, exactly via the amplification law $N_n=N_0(1+E)^n$ solved for $N_0$ .

Does PCR distinguish DNA from RNA, or living cells from dead ones?

PCR by itself amplifies DNA only; RNA targets require an added reverse-transcription step (RT-PCR) to first make complementary DNA. Standard PCR also cannot distinguish DNA from live versus dead cells, since it will amplify any intact target sequence regardless of whether it came from a viable organism — a real limitation in food safety and clinical microbiology, addressed by variants such as viability PCR that use DNA-intercalating dyes to selectively suppress amplification from DNA inside membrane-compromised (dead) cells.

Why do PCR primers need to be roughly 18–25 nucleotides long?

Primer length balances two opposing requirements. It must be long enough that its sequence is statistically unique within the genome being amplified — a 16-nucleotide sequence has a $4^{16}\approx4.3\times10^9$ chance combinatorial space, comparable to genome size, so shorter primers risk binding many unintended sites. It must also be short enough that its melting temperature stays in a practical, easily controlled range (roughly 55–65 °C) and that primer synthesis remains cheap and efficient. The 18–25 nt window is the empirical sweet spot satisfying both constraints simultaneously.

Why is Taq polymerase's lack of proofreading sometimes a feature rather than a flaw?

Proofreading exonuclease activity, while it improves fidelity, also tends to degrade single-stranded primers and can remove properly incorporated nucleotides from a blunt or recessed 3′ end when a polymerase pauses, complicating certain applications. Taq polymerase's simple lack of a 3′→5′ exonuclease domain makes it robust, fast, and highly processive, and its terminal transferase activity (adding a single non-templated adenine to the 3′ end of products) is actually exploited deliberately in TA-cloning, where PCR products with 3′-A overhangs are ligated directly into complementary 3′-T vector overhangs without any additional enzymatic modification step.

The Polymerase Chain Reaction (PCR)

Where this lives today

The physical picture — thorough prose

The core ideas — governing equations

Worked examples

Limiting cases, scope & common traps

The chemist behind it

Before this

Where this leads