Genome-Wide RNAi Screening to Identify Human Host Factors Crucial for In fl uenza Virus Replication - RNA Interference from Biology to Therapeutics - page 248

screens are initially unbiased, the necessity to establish criteria by which to select
hits automatically introduces bias: Scores and hit lists are dependent on experi-
mental setup of the screening systems employed (cell types, virus strains, reporter
assays, and variability among them) as well as on data evaluation methods, i.e., by
way of different analysis algorithms, Z-score cutoffs, and further in- or exclusion
criteria (see Table 12.1 ). Moreover, knockdown efficiency can vary among different
siRNAs used and can also be due to different protein expression levels and stabili-
ties among cell types. In addition, the screens cannot differentiate among factors
with redundant functions; these functions, thus, stay unaffected by single-gene
knockdown.
Several cross-comparative reviews have attempted to find common patterns
among the host target genes identified from the results of the genome-wide screens
discussed [ 5, 6, 28 ]. A very comprehensive, comparative review was performed by
Watanabe et al. which identified a total of 1,449 human genes potentially involved
during the influenza A infection and replication process. Watanabe et al. then nar-
rowed down this number to 128 candidates by pair-wise comparisons of the respec-
tive candidate genes and by selection of those found in at least two screens. They
include a screen by Sui et al. [ 29 ] into their bioinformatic assessment which employs
the so-called random homozygous gene perturbation (RHGP) technique based on
insertion of a lentiviral genetic element at a single site in one allele of the genome
in either the sense or antisense orientation. Unlike RNAi screens, the RHGP system
can theoretically knock down or overexpress any gene without any prior knowledge
or annotation of that gene, resulting in alteration of the phenotype.
In their metareview, Watanabe et al. found the number of genes common between
pairs of screens to be relatively small (from 0 to 32 genes in the 15 pair-wise compari-
sons performed). This low incidence of overlap between hits of the different screens
is similar to that observed in three RNAi screens aimed at identifying factors required
for HIV replication [ 30- 32 ]. In both cases, the most likely explanation for this low
degree of overlaps originates from differences within the screening systems, such as
cell type, virus strain, siRNA libraries employed, and/or individual readout system
that could result in bias among the screens. This explanation is supported by the fact
that the number of common hits among the IAV screens was highest among those
performed by König et al. and by our group. Both of these screens were performed
using the same siRNA library and both used the same human epithelial cell line.
Watanabe et al. use the PANTHER Classification System [ 33 ] as well as analysis
by Reactome [ 34 ] to assess for enriched gene categories among the 128 candidate
genes. These map the influenza virus-host interaction and respective enriched gene
targets to individual steps of the influenza virus life cycle, based on the known func-
tions of the identified host proteins. The greatest overlap among the hits discovered
in the individual screens is found among the following four groups of host factors:
(a) coating on vesicles-mediating retrograde transport from the Golgi to the ER and
endosomal trafficking mediated by coatomer 1 (COPI) transport complex ( ARCN1 ,
COPA , COPB2, and COPG ), (b) genes encoding the vacuolar ATPases (i.e.,
ATP6V0C , ATP6V0D1 , and ATP6V1A ), (c) nuclear transport of mRNAs and
proteins (i.e., KPNB1 , Nup98 , and NXF1 ), and (d) splicing of cellular pre-mRNA
(i.e., PTBP1 , NHP2L1 , SNRP70 , SF3B1 , SF3A1 , P14 , and PRPF8 ) [ 5 ] .