In the present study, we investigated associations of 70 index SNPs in 67 breast cancer susceptibility loci identified to date in up to 1,231 cases and 2,069 controls of AA women. We found that seven SNPs were significantly associated (P<0.05) and three SNPs were marginally significantly associated (P<0.10) with overall breast cancer risk in the same association direction as previously reported. Three additional SNPs showed a significant association (P<0.05) when stratified by breast cancer subtype. GRS analyses showed significant associations with the risk of overall or subtype of breast cancer.
In the present study population, on average, approximate 83% of genetic ancestry is African origin, which is similar to the estimate in other studies 
. Women in the SCCS have a higher African ancestry level than those in the NBHS. In the NBHS, most women were recruited in Tennessee, while SCCS women were recruited in 12 southern states including Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Tennessee, Mississippi, South Carolina, North Carolina, Virginia, and West Virginia. We did not find a significant difference in African ancestry level between breast cancer cases and controls or across breast cancer subtypes. These results are consistent with the recent finding in the CARE Study 
. However, in another study, significant association was observed between genetic ancestry with ER+, PR+, or localized breast tumors 
. In the present study, data for ER, PR and HER2 were available for only a portion of the subjects. Therefore, statistical power in regard to subtype of breast cancer analyses may be limited.
Among the ten SNPs that showed significant or marginally significant association with overall breast cancer risk in the present study, six have been investigated in previous studies of African-ancestry populations 
. SNP rs10069690 (5p15/TERT
) was originally discovered in a GWAS of AAs 
, so it is expected that this SNP should be replicated in the present study. For SNP rs1219648 at 10q26/FGFR2
, association was observed in our previous study 
, the Carolina Breast Cancer Study (CBCS) 
, and the Women’s Health Initiative (WHI) study 
, but not in the Women’s Insights and Shared Experiences study 
. SNP rs13387042 was replicated in our previous study 
and in a consortium study of 3,016 cases and 2,745 controls 
, but not in the WHI study 
, the CBCS 
, nor another pooled study 
. SNP rs8170 at 19p13/BABAM1
was only investigated in a pooled study from Africans and AAs and no association was observed 
. Both CASP8
SNP (rs1045485) and RAD51L
SNP (rs999737) have a MAF <5% in African-ancestry populations. Therefore, it is not surprising that these two SNPs were not replicated in all previous studies of African-ancestry populations 
. The other four SNPs that showed associations with overall breast cancer in the present study, rs4849887, rs17817449, rs13329835 and rs4808801, were recently discovered 
and have not been evaluated in previous studies of African-ancestry populations.
We did not replicate associations for the other 60 reported SNPs with the risk of overall breast cancer. Inconsistent results were reported for some of them in previous studies of African-ancestry populations. For example, SNP rs3803662 at 16q12/TOX3
was replicated in the WHI 
, but not in the Black Women’s Health Study (BWHS) 
, the CBCS 
and others 
. In addition, a significant association has been identified for this SNP with association in the opposite direction as previously reported 
. SNP rs2981582 (10q26/FGFR2
) was significantly associated with breast cancer risk in two studies 
, but not in the other studies 
. The WHI study reported a significant association for rs10941679 at 5p12/MRPS30
, and one study showed association for rs865686 at 9q31/KLF4
, however, other studies did not replicate these two SNPs 
In general, our results that approximately 90% of index SNPs were not replicated in AAs are consistent with results from previous studies in African-ancestry populations 
. Rebbeck et al 
did not find any association for three investigated SNPs. Huo et al 
evaluated 19 genetic loci and none of them were replicated. Five of the seven investigated SNPs in the CBCS 
, 21 of the 22 investigated SNPs in the WHI 
, and 14 of 19 SNPs in a consortium study 
were not replicated.
Because of the large difference in genetic architecture between African-ancestry and European/Asian ancestry populations, failure to replicate most of the reported SNPS in AAs is not surprising. Most, if not all, index SNPs identified in GWAS are associated with breast cancer risk through their strong LD with causal variants. African-ancestry populations have shorter LD and more genetic variations than European/Asian ancestry populations and may have different SNPs in LD with the causal variant. This may be the major reason why index SNPs are not replicated in African descendants. For example, in the BWHS, originally reported index SNPs rs10941679 and rs3803662 were not replicated, but other SNPs in these regions, rs16901937 and rs3104746, were associated with breast cancer 
. It has been reported that other markers were identified in AAs to better capture the association signal than the index SNPs originally discovered in the 2q35/TNP1
. Second, allele frequencies for the index SNPs differ considerably across ethnic groups. Many index SNPs have lower MAF in AAs than in Europeans/Asians. Even if the effect size of the index SNP is the same across populations, larger sample size is required to detect association in AAs due to the lower MAF. Third, the vast majority of SNPs were originally discovered among European descendants, who have a much higher proportion of ER+ than ER- breast cancer. Because of this, most of the reported risk variants are, in general, more strongly associated with ER+ than ER- cancer 
. African-ancestry women have a higher proportion of ER- breast cancer than European-ancestry women; this may be another reason for the non-replication in AAs.
To our knowledge, this is the first study in AAs that has evaluated index SNPs in all breast cancer susceptibility loci identified to date. However, the sample size in our study is relatively small, especially when stratified by breast cancer subtype. Some of the null associations observed in this study could be due to inadequate statistical power. Meta-analysis by pooling together all existing data in the AA populations will increase the statistical power to evaluate the effects of these variants in AAs. The other limitation of this study is that we only investigated index SNP in each locus. Large-scale fine mapping studies are needed to identify genetic risk variants at these loci in African-ancestry populations. Such work will be very helpful to identify causal variants for breast cancer.
In summary, in this African-ancestry population study, we replicated approximately 10% of index SNPs in 67 breast-cancer susceptibility loci. Heterogeneity was observed across the breast cancer subtype. These results show the complexity in applying GWAS findings to African-ancestry populations. Large-scale studies in AAs are needed to discover genetic risk variants which impact this population.