Fig 1.
Proposed mechanism for the creation of double-strand DNA breaks resulting from the interaction of widely-spaced inverted repetitive sequences.
A) Representation of an inverted repetitive sequence. Human inverted Alu element pairs are statistically depleted relative to direct-oriented Alu pairs up to a spacer size of 421,000 bp. B) DNA looping can create alignment of inverted pairs. C) Reactive single-stranded DNA is created both by DNA breathing bubbles as shown in this pane and also by replication forks. Note that the bases flip outward in breathing bubbles because of steric hindrance. This flipping out of bases makes the DNA more susceptible to interacting with a complementary breathing bubble. D) If single-stranded conformations of DNA occur in aligned inverted sequences, ectopic invasion and annealing can occur. E) Eight short stretches of DNA would be created as DNA makes the transition from its normal double-stranded form into a Doomsday Junction conformation. As with DNA hairpins, these short stretches of single-stranded DNA are susceptible to DNA nuclease attack (58). Random cleavage of the eight single-stranded DNA sites is hypothesized to continue until the Doomsday Junction is resolved. If both strands of the same DNA segment are cleaved, a double-strand break can be created.
Fig 2.
Estimated stability of overlapping 15 bp windows across EGFR Exon 19 of the hg38 reference genome assembly.
Note that while this exon is comprised of 99 bp, the X axis covers only 85 bp. This is because the last 15 bp window extends from base number 85 through base number 99. The canonical exon 19, 15 bp deletion (amino acids ELREA) extends from nucleotides 52–66. Finally, the stability of this cancer-linked region lies within the top 20% of the most unstable 15 bp regions within exon 19. This finding suggests that some of these regions of higher instability may not create driver mutations when damaged.
Fig 3.
Estimated stabilities of EGFR exon 19 for five genomic phases with overlapping 15 bp windows.
The increment for these windows is one base pair. Note that while these five phases present similar estimated stabilities, they are not identical. The blue arrows identify exon 19 regions where the stabilities are most dissimilar. As with Fig 2, the X axis covers only 85 bp. This is because the last 15 bp window extends from base number 85 through base number 99.
Table 1.
Reverse complement variation among five EGFR landscapes.
Fig 4.
Loci of reverse complement sequences to the locus of the EGFR canonical exon 19 deletion (hg38).
A) Total of 94 complementary sequences of 60% homology or greater are shown as red circles and fall within the reactive distance for inverted Alu element pairs. The Y axis is the number of mismatches found for each reverse complement located within the canonical deletion landscape. The maximum reactive distance of ± 421,000 bp is represented as black dashed lines [35]. Note that the most homologous locus to the canonical deletion is also the closest. This highly homologous locus accounts for almost 90% of the calculated instability of the exon 19 canonical deletion sequence. B) 50X magnified view of the canonical deletion landscape within ± 10,000 bp of the EGFR exon 19 canonical deletion locus.
Fig 5.
Reverse complement loci within ± 500,000 bp landscapes across five partially phased regions surrounding the EGFR exon 19 canonical deletion.
This mega-base sized landscape spans from ± 500,000 bp of the EGFR exon 19 canonical deletion and is represented across five panes, each showing 200,000 bp. Although each of the five genome sequences in Fig 5 are identified on the Y-axis, the phase designations for each proband sequenced in this study are arbitrary as several low coverage phase discontinuities (see discussion) exist along each of these four landscapes. Horizontal red lines in each pane identify phased regions within each landscape. Horizontal blue lines identify regions where the depth of sequencing is 10-fold or greater. The absence of heterozygosity across some phase blocks prevented complete phasing of the landscapes. These regions of low heterozygosity (non-phased) are represented by blue lines only, without red lines. Note that the locus of the canonical deletion is located in the center-most pane at the X-axis value of “0” (blue line). Also note the very close proximity of the highest homology reverse complement to the canonical deletion locus which is designated as red circles. This reverse complement is separated by only 1,366 bp from the canonical deletion sequence. This is the closest locus among all 94 sets of reverse complements to the EGFR exon 19 canonical deletion sequence and are within the ± 421,000 bp region of reactivity flanking the canonical deletion. This region of reactivity is set by the statistical confidence of departure of the I:D ratio of Alu elements departure from unity (p<0.05) (41). These statistical confidence limits are shown as blue dashed lines at ± 421,000 bp from the canonical deletion locus. Except for the solid red circles denoting the high homology reverse complement, all remaining reverse complements are identified as black circles with the number of mismatches from the 15 bp canonical deletion shown above each circle. Each red X denotes a missing reverse complement that is present in one or more of the other landscapes. A total of 22 reverse complements are absent across one or more of these five haplotypes shown in this Fig 5. It is important to note that although phase identity is not possible within each proband, the presence of homologous phase blocks across the probands permitted identification of reverse complement variation across these five ± 500,000 bp landscapes.
Fig 6.
Hypothesized mechanism for canonical driver mutation in EGFR exon 19.
A) Schematic of the region of both the EGFR exon 19 deletion prone region and its most homologous reverse complement. In this actual example from hg38, this reverse complement is located 1,366 bp 3’ from the EGFR exon 19 canonical deletion sequence. B) Alignment of these two sequences via formation of a 180° loop of DNA. C) Ectopic invasion and annealing (see also panes C, D, and E of Fig 1) through either aligned DNA breathing bubbles or replication forks. Note that the initial alignment is not perfect, and a single-base non-aligned region can occur D) Formation of Doomsday Junction. Putatively, this DNA conformation may be resolved through one or more nicks by single-strand nucleases at the transitional boundaries of this ectopic DNA conformation (see yellow lightning symbols). E) Regions of the inverted pair that are vulnerable to nicking. Comparing panes D and E, note that if a nick occurs in each of the two single-strands of DNA at the same end of a repetitive sequence, a double-strand break is created. If two single-stranded nicks occur at both ends of the exon 19 strand, the canonical deletion can be created.