Page 75 - Read Online

Page 6 Moore et al. J Transl Genet Genom 2021;5:200-217 https://dx.doi.org/10.20517/jtgg.2021.08

study), fraction of missing SNPs, and ten principal components of ancestry. Results were combined across
[31]
GWAS using METAL , and multiple-testing adjustment was performed using a Bonferroni correction.
RESULTS
Table 1 presents summary statistics for the ROH, FROH, and F3 by GWAS. Among participants, the
median total length of ROH ranged from 11,535 to 23,014 kb depending on the GWAS. The median
number of ROH per individual ranged from 4 in the UCSF2 GWAS, which used an older and less dense
GWAS chip, to 8 in the GEC GWAS, which included familial CLL cases. Median FROH ranged from 0.38%
-3
-4
to 0.77% of the autosome. Median F3 ranged from -6.99 × 10 to 3.02 × 10 [Table 1].
We discovered a positive association between the risk of CLL and increased homozygosity as measured by
-5
-6
FROH (β = 21.1, 95% SE = 4.41, P = 1.6 × 10 ) and F3 (β = 27.5, SE = 6.51, P = 2.4 × 10 ) [Table 2] with
limited evidence of between-study heterogeneity (P = 0.42 and 0.11, respectively) [Supplementary Table 3].
het
As CLL is an indolent lymphoma and there is a potential for tumor DNA contamination in the blood drawn
for genotyping, we performed a sensitivity analysis using only CLL cases and controls from prospective
nested case-control studies (ATBC, CPS-II, EPIC, HPFS, MCCS, NHS, NYU-WHS, PLCO, and WHI) from
the NCI NHL GWAS, where the DNA was often collected many years prior to diagnosis. Despite the
reduction in the number of CLL cases (n = 2140 to 889), the estimated association parameters for FROH
-3
and F3 were similar (FROH: β = 21.3, SE = 7.76, P = 6.04 × 10 ; F3: β = 22.5, SE = 5.45, P = 3.57 × 10 ).
-5
Figure 1A shows the P-values (-log ) from the meta-analysis assessing the associations of CLL with ROH
10
centered in each 500 kb bin in the autosome. No bins reached statistical significance after correction for
multiple testing (CLL Bonferroni alpha level = 0.05/45,590 bins = 1.1 × 10 ); the most significantly
-5
associated bin was located at chromosome 22q12.2 (P = 5.92 × 10 ). However, one of the top ten associated
-4
bins overlapped with the chromosome 13q14 region, a region where a somatic deletion is often seen in
[32]
CLL . To test whether the association between FROH and CLL was due to mosaicism at 13q14, we
excluded the 45 CLL cases identified to have this deletion from the NCI NHL GWAS. After removal of
these 45 cases, the association between FROH and CLL was slightly attenuated but remained statistically
-4
significant (β = 16.34, SE = 4.74, P = 5.71 × 10 ), similar to the results from the entire NCI NHL GWAS for
CLL (β = 18.9, SE = 4.76, P = 6.73 × 10 ).
-5
As structural alterations in chromosomes, such as trisomy 12, are a hallmark of CLL and could potentially
be mistakenly called as ROH using this method, we conducted a sensitivity analysis excluding chromosomes
12 and 13, which are frequently altered in CLL, from the meta-analysis . After excluding these two
[33]
chromosomes, the results for the association between FROH and CLL were found to be similar (β = 19.3,
SE = 4.4, P = 9.0 × 10 ) to previous results for all 22 autosomal chromosomes. Similarly, for F3, exclusion of
-6
chromosomes 12 and 13 led to results that are similar in magnitude and remain statistically significant (β =
17.6, SE = 7.4, P = 0.02). Between-study heterogeneity (I = 62.2%, P = 0.05) was detected in this sensitivity
2
het
analysis, largely attributable to the GEC GWAS.

We also identified positive associations between FL risk and increased FROH (β = 11.4, SE = 5.82, P = 0.02)
but not F3 (β = 13.2, SE = 8.01, P = 0.10) [Table 2]. There was evidence of between-study heterogeneity in
2
the meta-analysis of F3 and FL (I = 64.2%, P = 0.04) due to the USCF1/NHS GWAS, but not in the meta-
2
analysis of FROH and FL (I = 5.3%, P = 0.37) [Supplementary Table 3]. We performed a sensitivity analysis
using only cases and controls collected from prospective nested case-control studies (ATBC, CPS-II, EPIC,
HPFS, MCCS, NHS, NYU-WHS, PLCO, and WHI) as part of the NCI NHL GWAS. This resulted in a
reduction of the number of cases (n = 2085 to 521) and statistical power, but the association between FROH