PCR primer design basics: length, GC content and melting temperature

What a primer has to do before it can be judged

A PCR needs two primers because DNA polymerase can only extend an existing 3' end. The forward primer pairs with the bottom strand at the left edge of the target, the reverse primer pairs with the top strand at the right edge, and the region between them is copied. Addgene's primer design page adds two conditions: the 3' end of each primer has to match the template completely so extension can proceed, and the two primers have to be oriented so their 3' ends point toward each other.

Every numeric guideline below is a proxy for two competing needs. The primer must bind its intended site firmly enough at the annealing temperature to be extended, and it must bind nothing else firmly enough to be extended. Length, GC content and Tm all trade one need against the other, which is why published ranges are guidelines with overlap rather than hard limits.

Primer length: the sources give overlapping ranges

Addgene lists a length of 18 to 24 bases. New England Biolabs, in its guidelines for Taq DNA polymerase, says primers are generally 20 to 30 nucleotides. The two ranges overlap at 20 to 24, and neither is a rule of nature. Addgene adds the failure modes at each extreme: short primers give inaccurate, nonspecific product, and primers longer than about 30 bases hybridize more slowly.

A back-of-envelope calculation shows why short is risky. In a random sequence of 3 billion bases, counting both strands, a particular 12-base string is expected about 358 times, a particular 18-base string about 0.09 times, and a particular 20-base string about 0.005 times. Real genomes are not random, with repeats and skewed composition, so treat these figures as an illustration of how fast specificity grows with length, not as a prediction for any organism.

Length also feeds directly into Tm, because a longer primer has more base pairs to hold it on the template. That is why length, GC content and Tm are best tuned together instead of one at a time.

GC content, even spacing and the 3' end

Both Addgene and NEB give 40 to 60 percent G plus C as the target range. A G-C pair has three hydrogen bonds and an A-T pair has two, so GC-rich primers hold more tightly at a given length. Because each base pair in a low-GC primer contributes less, such a primer needs more length to reach the same Tm, which is one reason the GC window and the length range are chosen together.

NEB also advises spacing G and C residues evenly within the primer. A primer with all its G and C bases bunched at one end and a long AT-only stretch at the other has a lopsided binding profile: one half anchors, the other half melts off easily.

For the ends, Addgene says primers should start and end with one or two G or C bases. This is often called a GC clamp, and it applies mostly to the 3' end, since that is the end the polymerase extends. Neither source puts a number on how many G or C bases at the 3' end are too many, so this article does not invent a limit.

Three ways to estimate Tm, and which one this site uses

The melting temperature is the temperature at which half of the primer-template duplexes are paired and half are separated. SantaLucia's 1998 paper defines it that way and gives it as a function of the duplex's enthalpy and entropy change plus the strand concentration. Every calculator is an approximation of that quantity, and the approximations differ in how much they know about the sequence.

The Wallace rule counts bases: 2 degrees for each A or T and 4 for each G or C. The basic GC formula in the site's code is 64.9 + 41 x (G+C - 16.4) / length. Both ignore which bases are neighbors and both ignore salt and primer concentration. The code comments limit them to short primers under 14 bases and to primers over 13 bases, respectively.

Nearest-neighbor thermodynamics treats the duplex as a chain of overlapping two-base steps. SantaLucia's unified parameters give an enthalpy and entropy value for each of the ten distinct Watson-Crick doublets, plus separate initiation terms for a duplex ending in G-C versus A-T. The primer Tm calculator on this site computes all three. Its nearest-neighbor estimate uses those unified parameters, SantaLucia's monovalent-salt entropy correction, a default of 0.25 micromolar primer and 50 millimolar sodium, and it returns no value for sequences containing ambiguity codes.

One simplification is worth knowing. The code divides the strand concentration by 4, which is the form the paper gives for non-self-complementary strands at equal concentration. In a real PCR the primer is in large excess over template, so this is an approximation, and other calculators make different choices there.

sequence              nt   GC%   Wallace  GC formula  nearest-neighbor
GCGATCGCAT            10   60.0  32.0     22.3        34.4
GTAAAACGACGGCCAGT     17   52.9  52.0     47.1        51.6
ATGGCTAGCAAAGGAGAAGA  20   45.0  58.0     49.7        52.5

Why two calculators can disagree by several degrees

The table above shows the formulas diverging as the primer gets longer: at 20 bases the Wallace figure is 5.5 degrees above the nearest-neighbor one, and the GC formula is 2.8 below it. Even two nearest-neighbor calculators can differ, because they may use different parameter sets, different salt corrections and different assumptions about strand concentration.

Conditions matter as much as the formula. In SantaLucia's salt analysis, a tenfold change in sodium concentration shifts Tm by about 11.7 degrees for oligonucleotides. The paper reports an average deviation of 2.2 degrees between its predictions and 26 measured oligonucleotide duplexes in 0.01 to 0.3 molar NaCl, so even in the conditions it was fitted for, a nearest-neighbor Tm is a good estimate rather than a measurement.

The site's calculator models only monovalent cations and primer concentration. It does not model magnesium, which real PCR buffers contain, or additives. For that reason NEB points users to its own Tm calculator to choose an annealing temperature. Treat any generic number as a starting point for a test, not a final setting.

sodium 20 mM  -> 48.0 C
sodium 50 mM  -> 52.5 C
sodium 100 mM -> 56.0 C

primer 0.10 uM -> 51.2 C
primer 0.25 uM -> 52.5 C
primer 0.50 uM -> 53.5 C
primer 1.00 uM -> 54.5 C

Matching the pair and choosing an annealing temperature

Both primers are present in the same reaction and share one annealing step, so their Tms should be close. NEB and Addgene both give a limit of 5 degrees between the two primers. A larger gap means one primer is either binding loosely or binding to sites it should not at the temperature that suits the other.

For the annealing step, Addgene sets it 5 degrees below the primer Tm, and NEB says typical annealing temperatures are 5 degrees below the lowest primer Tm and often fall between 50 and 60 degrees C for Taq. If the gel shows spurious products, both sources say to raise the temperature: NEB says to test higher annealing temperatures, and Addgene says to increase it stepwise by 1 to 2 degrees.

The example below applies those rules to a pair written for this article. Each Tm is close to the other, but the reverse primer is below the 40 percent GC floor and both primers end in A rather than G or C, so a designer following the Addgene guidelines would slide the primer positions a few bases and recalculate. The starting annealing temperature also falls under NEB's 50 to 60 range, though NEB's acceptable calculated Tm range of 42 to 65 still includes both primers.

template 5'-ATGGCTAGCAAAGGAGAAGAACTTTTCACTGGAGTTGTCCCAATTCTTGTTGAATTAGATGGTGATGTTAATGGGCACAAATTTTCTGTC-3'

forward  ATGGCTAGCAAAGGAGAAGA       20 nt  45.0% GC  Tm 52.5 C
reverse  GACAGAAAATTTGTGCCCATTA     22 nt  36.4% GC  Tm 51.6 C

Tm difference           : 0.91 C
lowest Tm minus 5       : 46.6 C

Writing the reverse primer: reverse and complement

The reverse primer is the part beginners most often get wrong. The two strands of DNA run in opposite directions, and primers are conventionally written 5' to 3'. To get the reverse primer, take the last 18 to 30 bases of the top-strand target region, complement each base, and then reverse the order so the result reads 5' to 3'. That is the reverse complement.

Complementing without reversing gives a sequence that pairs with the template but reads 3' to 5', so if it is ordered as written it will not match the intended site. In the example, the last 22 bases of the template are TAATGGGCACAAATTTTCTGTC, and reverse-complementing them gives GACAGAAAATTTGTGCCCATTA, which is the reverse primer in the previous section. Applying the operation twice returns the original, a quick self-check.

top strand, last 22 bases : TAATGGGCACAAATTTTCTGTC
reverse complement        : GACAGAAAATTTGTGCCCATTA
complement only (wrong)   : ATTACCCGTGTTTAAAAGACAG
reverse complement of the reverse complement equals the original: true

Hairpins, primer-dimers, runs and repeats

A primer can pair with itself. A hairpin forms when one stretch of the primer is complementary to another stretch of the same primer, and a primer-dimer forms when two primers anneal to each other instead of the template. Addgene describes dimers as disrupting amplification and says primer pairs should not have complementary regions, and NEB says to avoid hairpins within each primer and dimerization between the two.

The 3' ends deserve the most attention. Because polymerase extends any paired 3' end, a dimer with a paired 3' end can be extended, and the product then competes with the target for reagents. This follows from how the enzyme works, and it is why complementarity at the 3' end is usually treated as worse than complementarity at the 5' end.

Long runs of one base and short repeated units, such as a dinucleotide repeated many times, give a primer many nearly equivalent ways to pair with itself or with the template at shifted positions. The sources used here give no run-length cutoff, so none is stated. A practical check is to reverse-complement one primer and look for it inside itself or inside the partner primer, for example with the pairwise alignment tool.

primer             : ACGAATTCGT
reverse complement : ACGAATTCGT
identical: true, so two copies of this primer can pair along their full length

Specificity: checking against the genome, not just the target

Everything above concerns the primer and its intended site. It says nothing about whether the sequence occurs somewhere else in the sample. NCBI's Primer-BLAST addresses this: it designs primers with Primer3 (version 2.5.0), then uses BLAST and a global alignment algorithm to screen them against a database the user selects, in order to avoid pairs that can cause nonspecific amplification.

It checks all combinations of primers, not only the intended forward and reverse pair. Forward with forward and reverse with reverse are screened as well, because one primer that binds two sites on opposite strands, facing each other, can make a product on its own. The database choice matters, because a primer can be unique in one collection of sequences and not in another.

None of the calculators on this site search a sequence database, so the tools here cover composition, Tm and orientation. Specificity needs a search against a real database, and that step should come after the primer has passed the simpler checks.

Worth knowing. A primer that passes every number in this article can still amplify the wrong product. Length, GC content and Tm describe the primer; only a database search describes the sample.