Page 80 - Read Online

Moore et al. J Transl Genet Genom 2021;5:200-217 https://dx.doi.org/10.20517/jtgg.2021.08 Page 11

[43]
population stratification . Even after adjustment using principal components, some population
substructure may have affected the GELA/EPIC study, leading to results that differ from the other DLBCL
studies. Our analysis of FL similarly included one GWAS that combined the cases and controls from 2
different United States studies. Although we did observe some heterogeneity among the FL studies for F3,
we did not observe such heterogeneity for FROH and FL, suggesting the F3 may be more sensitive to
population substructure than FROH. The patients in the UCSF1/NHS study were slightly younger and less
likely to be female than the cases in the NCI GWAS, which could have also contributed to the heterogeneity
in results.

DLBCL is known to have substantial disease heterogeneity [44,45] . It is possible that a greater understanding of
the heterogeneity of DLBCL and a further molecular subtype-specific analysis, which was not possible for
the present study, may allow a better elucidation of any association between ROH and disease risk. While
the analysis of F3 and DLBCL was similarly affected by between-study heterogeneity [Supplementary Table
3], a potential advantage of F3 over FROH as a measure of homozygosity is its expected smaller variance
and bias . In addition, the method of calculating FROH used in this study does not count homozygosity
[30]
[28]
below a length threshold of 1500 kb . It is possible that a portion of the recessive genetic variation in
DLBCL resides in short, common ROH of ancient ancestry that was unmeasured by our approach.
[14]
Although family history of NHL or other hematologic cancers has been associated with increased risk of
MZL , we did not find evidence linking FROH or F3 to MZL. Our published GWAS of MZL found two
[46]
[11]
independent loci conferring risk, both in the HLA region of the genome , and a follow-up study reported
that homozygosity at class I HLA-B and -C and class II HLA-DRB1 loci was associated with increased risk
of MZL . Although our study did not detect an association with homozygosity more broadly across the
[41]
genome, local homozygosity, at least in select regions, is still likely to play a role in risk. Of the four subtypes
of NHL examined in the present study, MZL is the least common. Our power to detect an association with
MZL was limited due to the small sample size, which was possibly compounded by the known heterogeneity
within MZL, with multiple recognized subtypes . Given that we expect ROH to capture the effects of small,
[47]
scattered, recessive genotypes, a larger sample size is likely needed to elucidate the genetic etiology of MZL.
As other highly heritable traits, such as certain autoimmune diseases and atopy, are associated with the risk
[46]
of MZL , identifying shared genetic architecture with these more common phenotypes may be a
complementary strategy to further elucidate the underlying genetic architecture.

This study had both limitations and strengths. The use of FROH as a measure of homozygosity is known to
[30]
require large sample sizes . We were able to combine eight GWAS across four subtypes of NHL to increase
our sample size and provide a fairly comprehensive analysis of recessive inheritance. However, as discussed,
we may have had insufficient sample size to detect associations for MZL. Further, we could not separately
examine clinically relevant molecular subtypes of DLBCL. ROH patterns are exquisitely sensitive to
[48]
population history . Thus, we restricted our study population to individuals of European ancestry and
adjusted for principal components of ancestry. This limited the potential for problematic population
stratification in our study sample, but it also reduced the generalizability of the findings to individuals of
other ancestries. To date, the optimal approach to analyze specific ROH to identify genomic regions with
novel, recessively acting risk alleles has not been established, but both the ROH calling procedure and the
use of F3 are accepted as means of identifying recent relatedness in outbred European populations [28,30] and
can provide evidence for the presence of recessive loci. Although it is possible that the associations between
FROH and F3 and CLL are due in part to DNA contamination by tumor cells, our sensitivity analyses
suggest that germline homozygosity is likely an independent risk factor for CLL. Finally, although our study
provides clues as to the genetic etiology of NHL, the clinical value of these findings is uncertain. Additional