Page 79 - Read Online

Page 10 Moore et al. J Transl Genet Genom 2021;5:200-217 https://dx.doi.org/10.20517/jtgg.2021.08

removed chromosomes 12 and 13 from the calculation of FROH and F3, as these two chromosomes are
[33]
frequently altered in newly diagnosed CLL , and again saw results similar to the primary analysis. While
we cannot exclude the possibility that some of the association could be due to erroneous homozygosity
calling of structural variation, the robustness of the associations indicates that potential bias from these
sources probably does not explain the observed association.

We did not identify any specific homozygous regions that were associated with CLL risk after adjustment
for multiple comparisons, but we observed nominal associations with ROH at chromosomes 22q12.2 and
13q14.2. The chromosome 22q12.2 region includes POZ/BTB and AT hook containing zinc finger 1
(PATZ1), which is a transcription factor that interacts with p53 and may modulate p53-mediated
[37]
apoptosis . The region also contains phosphoinositide-3-kinase interacting protein 1 (PIK3IP1), which
[38]
inhibits antitumor immunity . The chromosome 13q14.2 region contains several genes of potential
interest, including RB transcriptional corepressor 1 (RB1) and cysteinyl-leukotriene receptor 2 (CYSLTR2).
[32]
RB1 is a known tumor suppressor, which is frequently somatically deleted in CLL . CYSLTR2 is part of the
leukotriene pathway and thought to play a role in lymphocyte proliferation and maturation .
[39]
For FL, we discovered a positive association between FROH and risk of FL, suggesting recessive genetic
variation is a contributor to risk. Previous epidemiologic work has shown that family history of NHL is a
[4,5]
risk factor for FL , and first-degree relatives have a 4-6-fold increase risk of FL . The largest GWAS of FL
[40]
to date detected multiple genetic loci associated with risk, with the strongest associations seen in the HLA
[10]
region . Consistent with our findings, a previous study of first-degree relatives suggested that FL may
follow an autosomal recessive mode of inheritance . In sensitivity analyses, we showed that results for
[5]
FROH are similar after excluding the entirety of chromosome 6, suggesting that our findings were not due
to the HLA region. This finding supports the role of non-HLA variation in the etiology of FL and provides
evidence that there may be additional rare, recessive loci that are associated with the risk of FL. We
observed similar results after limiting to cases with prospectively collected DNA, suggesting that tumor
contamination is unlikely. Although we did not find any individual regions (based on 500-kb bins of ROH)
that were significantly associated with the risk of FL after adjustment for multiple testing, we observed
nominal evidence for homozygosity overlapping the HLA region at chromosome 6p21.33. We previously
reported that homozygosity for classical HLA class II alleles, such as HLA-DRB1, was associated with an
increased risk of FL compared to individuals with heterozygosity . These findings are consistent with the
[41]
hypothesis that HLA homozygosity may increase risk by reducing an individual’s ability to recognize a
diverse set of foreign antigens.

DLBCL shows a strong association with family history of NHL in epidemiologic studies , and it has been
[42]
previously estimated that 16% of the variation in DLBCL is due to common genetic variants . Overall, we
[8]
did not observe an association between DLBCL and homozygosity, suggesting the recessive variation may
not have a strong role in DLBCL susceptibility, a finding consistent with a previous study of DLBCL among
first-degree relatives . The meta-analysis for DLBCL was affected by substantial between-study
[5]
heterogeneity (P = 0.01 for FROH and P = 0.001 for F3). In the leave-one-out meta-analysis, removing
het
het
2
either the NCI NHL GWAS or the GELA/EPIC GWAS reduced the I among the remaining three GWAS to
0%, and significant associations were observed with FROH and F3 after removing GELA/EPIC. The
GELA/EPIC GWAS was different from the other studies in that it included patients from clinical trials, who
were slightly younger and less likely to be female than the other studies [Supplementary Table 2]. In
addition, this GWAS combined cases from French clinical trials with controls from various European
countries in the EPIC cohort. It is possible that the degree of population matching, while adequate for a
GWAS, is not sufficient for analyses of F3 and FROH, which are known to be especially sensitive to