Information about the technology employed by Illumina is primarily taken from publications of the company.
In 2002 Illumina entered the microarray market by producing genotyping chips. Two years later the company launched expression profiling microarrays for man and mouse, serving both whole transcriptome as well as customised applications.
In general, there are three distinct approaches to manufacturing microarrays carrying oligonucleotides:
- Oligonucleotides are synthesized and spotted by an automated device
- DNA can be synthesized directly on the surface
- Oligonucleotides are synthesized and loaded onto discrete particles that are finally immobilized on chips
The first approach has been classically used to produce home-made oligonucleotide microarrays (e.g. Operon oligo sets).
The second approach provides wide flexibility in design of custom arrays without preliminary synthesis of modified oligonucleotides (e.g. Affymetrix, Nimblegen, febit).
However, in situ synthesis does not allow quality control and purification of individual array features, and incomplete phosphoramidite coupling prevents high yields of full-length oligonucleotide probes, especially for longer oligos.
Figure 1: Schematic view of a bead coupled with an oligonucleotide, consisting of the address code and a 50 base gene-specific sequence
The third approach uses presynthesized oligonucleotides. In detail, Illumina’s BeadArray™ technology uses silica microspheres (beads) as the array elements. Each microsphere of a diameter of 3 µm, is derivatized with a particular oligonucleotide that acts as a probe for the complementary sequence in an assay solution (Steinberg et al., 2004). The oligonucleotides synthesized are >72 bases, consisting of an address code and a 50 base gene-specific sequence. Each oligonucleotide is tightly quality controlled and coupled to a batch of beads (see figure 1).
Each bead carries >1x105 identical >72mer oligonucleotides. For a 47k gene expression array, 47,000 bead types are prepared, and equal aliquots of each type are pooled. The pool is spread across prefabricated microarrays with defined microwells that fit to the bead size (see figure 2). The beads are immobilized within the cavities and the 25 base addresses are decoded, to allocate each bead to the respective gene sequence (Gunderson et al., 2004). The error rate in decoding is considered low (<1x10-4).
Each bead type has an average 30x representations on the chip – a strategy that provides the statistical accuracy of multiple measurements.
- Figure 2: Production of BeadChips
Labeling procedure
- The total mature RNA is isolated from the cell/tissue being studied. This RNA has already been “processed” (removal of the noncoding introns and splicing together of the coding exon) as well as the addition of a poly-A tail
- The RNA is turned into a double stranded DNA copy known as a cDNA. This is done through reverse transcription. This is done because RNA itself is not a very stable molecule and the cDNA is a way to store the RNA for a much longer period of time
- When it comes time to run the array, the cDNA is allowed to go through in vitro transcription back to RNA (now known as cRNA), but this RNA is labeled with Biotin. This is done by having uracil bases tagged with the Biotin.
- The Biotin-labeled cRNA is then added to the array
- Anywhere on the array where a RNA fragment and an oligonucleotide on a bead are complimentary, the RNA sticks to the probe on the bead
- The array is then washed to remove any RNA that is not stuck to an array (i.e., no match was made) and then stained with the fluorescent molecule that sticks to Biotin
- Lastly, the entire array is scanned with a laser and the information is kept in a computer for quantitative analysis of what genes were expressed and at what approximate level.
The Sentrix BeadChips are always composed of multiple entities on one physical chip (see figure 3). Sentrix-6 BeadChips can analyze 47,000 genes on one chip. The genes to be analyzed are subdivided into a 23k set that corresponds to so-called RefSeq genes. These are genes (and probes) that are highly annotated, and mostly curated by the NCBI Reference Sequence project (see "a stripes" in figure 3). The "b stripes" represent a ~24k set that are made up from less well annotated genes, mostly defined by sequence clusters as defined by unigene (NCBI project).
As Sentrix-8 BeadChips lack the "b stripes", they are good for the analysis of 23k genes.
Due to the geometry of the Sentrix chips it is most economic to always run six probes on a Sentrix-6 chip (or 8 probes on a Sentrix-8 chip), as the chips cannot be subdivided or re-used.
- Figure 3: Sentrix BeadChips
The performance of BeadChips has been tested for their RNA dose response as well as for their dynamic range (see figure 4).
Known quantities of reporter gene RNAs that lack sequences present in human RNA were spiked into human RNA samples. The range over which differences can be measured was assessed by doping a series of samples with increasing concentrations of spiked genes. This dose-response curve was generated for each gene upon hybridization to the array. In total, eight artificial polyadenylated RNAs were generated in vitro from cloned bacterial and viral genes. Each of these RNAs was labeled and spiked at twelve different concentrations (0, 0.01, 0.03, 0.1, 0.3, 1, 3, 10, 30, 100, 300, 1000 pM) across a panel of twelve aliquots of labeled human cRNA. This background RNA was generated from Universal Human Reference RNA. Each spiked, labeled sample (1.5 µg per sample) was hybridized in quadruplicate across 48 arrays on eight different Human-6 Expression BeadChips. Each sample replicate was hybridized on a different BeadChip to avoid the possibility of BeadChip bias. The dose-responses seen and their variation across replicates provided the basis of calculating assay performance metrics.
Based on these curves, and the spread around each data point, the calculated median limit of detection for this experiment was 0.2 pM (99% confidence). The median dynamic range, defined as the concentration range over which 2-fold changes can be detected with 95% confidence, was measured as > 3 logs, and the median concentration change detectable with 95% confidence was measured as 1.3-fold.
- Figure 4: Dose-response curve
