Illumina Bridge Amplification: How Sequencing Clusters Form on a Flow Cell

What It Is and Where It Happens

Bridge amplification is the cluster-generation step of Illumina sequencing — a form of solid-phase PCR in which single library molecules are amplified into clonal clusters directly on the glass of a flow cell, without any free-floating template.

The flow cell is a thin glass slide with one or more channels (lanes). During manufacturing its surface is chemically coated ('grafted') with an ultradense lawn of two short single-stranded oligonucleotides, universally called P5 and P7, at a density of roughly 10⁶–10⁷ oligos per μm². Every DNA fragment in a prepared library carries an adapter whose ends are complementary to P5 and P7.

• Purpose: turn one molecule into ~1,000 identical copies so the fluorescent signal is bright enough to image.
• Clonality: because copies stay tethered near their origin, each cluster is monoclonal — all strands share one template sequence.
• Spatial separation: clusters are physically distinct spots, letting the instrument read millions of them in parallel.

The Mechanism, Step by Step

Cluster generation proceeds as a repeating isothermal cycle. A single denatured library strand flows across the lawn and hybridizes by its P5-complementary end to a surface P5 oligo.

• 1. Extension: Bst DNA polymerase extends the grafted P5 primer along the template, making a covalently anchored complementary copy.
• 2. Denaturation: formamide (not heat) melts the duplex; the original template washes away, leaving one tethered single strand.
• 3. Bridging: the free P7 end of that strand bends over and hybridizes to a nearby surface P7 oligo, forming the arch-shaped bridge.
• 4. Bridge extension: polymerase copies the bridged strand into a double-stranded bridge, now anchored at both ends.
• 5. Denaturation: formamide separates the bridge into two tethered single strands.

Repeating this ~28–35 times amplifies one molecule into a tight cluster of ~1,000 strands. Because each new strand tethers close to its parent, the cluster stays compact (~1 μm across).

Key Molecules and Characteristic Numbers

A few components do all the work, and their real specifications explain the platform's performance.

• P5 adapter (5'-AATGATACGGCGACCACCGA...) and P7 adapter (5'-CAAGCAGAAGACGGCATACGAGAT...) — the universal grafting/priming sequences; the numbers are historical part designations, not chemistry.
• Bst polymerase — a strand-displacing, thermostable Bacillus stearothermophilus enzyme that works isothermally near 60 °C, avoiding thermocycling on the surface.
• Formamide — chemical denaturant replacing heat, lowering DNA melting temperature so cycles stay isothermal.
• USER enzyme (Uracil-DNA glycosylase + Endonuclease VIII) — cleaves a deoxyuridine built into one primer to linearize strands before sequencing.

Characteristic numbers: ~2 billion clusters on a patterned NovaSeq flow cell; ~1,000 copies/cluster; ~35 cycles; oligo lawn ~10⁶–10⁷/μm²; whole step ~1–4 hours. Loading concentration is tuned (a few hundred pM) so ideally one molecule seeds each nanowell.

Linearization, Blocking, and How It's Controlled

After amplification, a cluster contains both forward and reverse strands — but you can only read one at a time cleanly, so the chemistry is groomed before sequencing begins.

• Linearization: the reverse (P5-anchored) strands are selectively cleaved and washed away. Illumina builds a cleavable site (a deoxyuridine or 8-oxo-G) into the P5 lawn oligo; USER enzyme or a glycosylase nicks it, releasing one strand type so a monoclonal forward template remains.
• 3' blocking: remaining surface oligos are dideoxy-capped so they cannot prime spuriously during sequencing.
• Primer hybridization: the Read 1 sequencing primer anneals, and sequencing-by-synthesis begins.

Quality is monitored by cluster density and %PF (passing filter). Overloading causes overlapping clusters the software cannot resolve; underloading wastes capacity. On patterned flow cells, nanowells impose a fixed cluster geometry, and 'exclusion amplification' (ExAmp) chemistry lets the first molecule to seed a well outcompete latecomers — improving monoclonality.

How It Compares to Related Processes

Bridge amplification solves the same problem as other clonal amplification methods — making one molecule bright enough to sequence — but does it in a distinctive way.

• vs. emulsion PCR (454, Ion Torrent): emulsion PCR amplifies onto beads inside oil droplets, then deposits beads into wells. Bridge amplification skips beads and droplets entirely, amplifying in place on flat or patterned glass — cleaner, but PCR-based, so it can generate duplicate reads and phasing.
• vs. rolling-circle amplification (MGI/BGI DNBSEQ): that platform circularizes templates and uses linear rolling-circle replication to make DNA nanoballs, avoiding exponential PCR bias — but bridge amplification's tethered geometry gives exceptionally uniform, dense clusters.
• vs. solution-phase PCR: ordinary PCR mixes primers and template freely, producing an amplicon soup; bridge amplification keeps every copy spatially anchored, which is what makes massively parallel imaging possible.

It is also distinct from the later sequencing-by-synthesis step it feeds: amplification builds the cluster; SBS reads it base by base with reversible terminators.

Significance, Applications, and Open Questions

Bridge amplification is one of the enabling inventions of the genomics era. It was described in Bentley et al., Nature 2008 ('Accurate whole human genome sequencing using reversible terminator chemistry'), building on Solexa technology acquired by Illumina, and it underpins the dramatic collapse in sequencing cost from ~$100 million per genome to under $1,000.

• Applications: whole-genome and exome sequencing, RNA-seq, single-cell sequencing, ChIP-seq, clinical oncology panels, pathogen surveillance, and metagenomics all depend on cluster generation.
• Disease relevance: reliable clusters underlie accurate variant calling; cluster artifacts (index hopping, PCR duplicates, phasing) can create false variants, so wet-lab and bioinformatic corrections matter clinically.

Open questions and limits: read length is capped (~2×250 bp) partly because phasing errors accumulate as clusters lose synchrony over cycles; PCR-based amplification introduces GC bias and duplicates that PCR-free library preps and unique molecular identifiers (UMIs) mitigate. Competing schemes (nanoball chemistry, single-molecule long-read platforms) address the amplification step's inherent biases from different angles.

Bridge amplification vs. other clonal amplification methods used in sequencing

Feature	Illumina bridge amplification	Emulsion PCR (454 / Ion Torrent)	Rolling-circle (MGI DNB)
Format	Solid-phase, on flow-cell surface	In water-in-oil droplets on beads	In-solution, then loaded to patterned array
Amplification chemistry	Isothermal bridge PCR, ~35 cycles	Thermocycled PCR in emulsion	Rolling-circle replication (linear, no exponential PCR)
Product per feature	~1,000-copy clonal cluster (~1 μm)	Bead coated with clonal copies	DNA nanoball (~200 nm concatemer)
Error mode	Low; PCR duplicates & phasing	Homopolymer errors, mixed beads	No PCR bias; needs patterned array
Key reagents	P5/P7 lawn, Bst pol, formamide	Beads, oil emulsion, Taq	Circularized template, phi29-type pol