24 2 Trackable Multiplex Recombineering (TRMR) and Next-Generation Genome Design Technologies
selective for the libraries and are grown to late log phase. An aliquot of this cul-
ture is frozen for further analysis, with the remainder of the culture being used
for selections. For selections in liquid medium, cells from the initial culture are
inoculated into medium containing the desired chemical compound and are
grown to stationary phase. Cells are then harvested for analysis as discussed in
the following section. For selections on solid medium, cells from the initial cul-
ture are spread on plates containing the desired chemical compound, and plates
are incubated until colonies are visible. All colonies are scraped from the plates
for further analysis.2.3.3 Analysis of Results
Either microarray or sequencing analysis can be used to determine the relative
fitness of each allele after selection. Genomic DNA is extracted from cells before
and after selection (and at various points during selection if desired), and the
molecular barcodes from each sample are amplified by PCR. In the original set of
TRMR experiments, barcodes were hybridized to the GenFlex Tag4 array from
Affymetrix [30]. A separate microarray experiment was done for each library.
Ten barcodes were spiked into the genomic DNA mixture in known amounts to
determine barcode concentrations, and 1642 negative-control barcodes were
also included to determine background hybridization rates. Allele frequencies
for each gene were then determined by dividing allele concentrations by the total
concentration of all alleles detected on the array.
In T^2 RMR, molecular barcodes that are optimized for high-throughput
sequencing were used instead of microarray to track alleles. Each allele had two
barcodes – one identifying the library (off, low, intermediate, high) and one iden-
tifying the gene. All samples were combined into a single MiSeq run. High-
throughput sequencing allows more quantitative analysis of genotype frequencies,
since individual alleles are directly tracked at the nucleotide level rather than by
relative hybridization intensity (measured in arbitrary fluorescence units). A sin-
gle run of Illumina MiSeq can generate 10^6 –10^7 sequencing reads (a typical
microarray signal distribution ranges over about 10^3 ), allowing for each barcode
to be sequenced thousands of times. This deep sequencing additionally aids in
the detection of rare alleles [31], which might be present in too low a concentra-
tion to be identified by microarray hybridization. A microarray hybridization
signal can also saturate [30], resulting in loss of data for the most highly expressed
alleles. A second advantage to high-throughput sequencing is that it results in a
lower error rate in identifying alleles. Although hybridization to a microarray, in
general, gives high fidelity, some alleles will fail to hybridize to the sequence on
the microarray that is perfectly complementary [32]. Furthermore, errors in bar-
code sequences can be introduced during cell replication, DNA synthesis, or
PCR amplification, resulting in loss of hybridization or, worse, in hybridization
to an incorrect spot on the microarray [32]. With high-throughput sequencing,
on the other hand, errors in barcode sequences can be identified and corrected
or discarded [33], as was done with T^2 RMR. In T^2 RMR, allele frequencies for
each gene were determined by dividing the number of barcode counts for that
gene by the total number of barcode counts for all genes.