Multi-ethnic transcriptome-wide association study of prostate cancer.

The genetic risk for prostate cancer has been governed by a few rare variants with high penetrance and over 150 commonly occurring variants with lower impact on risk; however, most of these variants have been identified in studies containing exclusively European individuals. People of non-European ancestries make up less than 15% of prostate cancer GWAS subjects. Across the globe, incidence of prostate cancer varies with population due to environmental and genetic factors. The discrepancy between disease incidence and representation in genetics highlights the need for more studies of the genetic risk for prostate cancer across diverse populations. To better understand the genetic risk for prostate cancer across diverse populations, we performed PrediXcan and GWAS in a case-control study of 4,769 self-identified African American (2,463 cases and 2,306 controls), 2,199 Japanese American (1,106 cases and 1,093 controls), and 2,147 Latin American (1,081 cases and 1,066 controls) individuals from the Multiethnic Genome-wide Scan of Prostate Cancer. We used prediction models from 46 tissues in GTEx version 8 and five models from monocyte transcriptomes in the Multi-Ethnic Study of Atherosclerosis. Across the three populations, we predicted 19 gene-tissue pairs, including five unique genes, to be significantly (lfsr < 0.05) associated with prostate cancer. One of these genes, NKX3-1, replicated in a larger European study. At the SNP level, 110 SNPs met genome-wide significance in the African American study while 123 SNPs met significance in the Japanese American study. Fine mapping revealed three significant independent loci in the African American study and two significant independent loci in the Japanese American study. These identified loci confirm findings from previous GWAS of prostate cancer in diverse populations while PrediXcan-identified genes suggest potential new directions for prostate cancer research in populations across the globe.

Introduction

In the past two decades, genome-wide association studies (GWAS) have evolved as a critical method to detect and characterize genomic loci affecting susceptibility to various polygenic disorders. As this important method to detect genetic associations has grown, individuals of non-European ancestries have made up less than 22% of all GWAS. Individuals of African, East Asian, and Latin American ancestries made up 2.03%, 8.21%, and 1.13%, respectively [1]. These populations are similarly poorly represented in GWAS of prostate cancer. Individuals of European ancestries make up over 85% of discovery GWAS, while individuals of African, East Asian, and Latin American ancestries make up less than 11%, 3%, and 1%, respectively [2]. Individuals of African American or Latin American ancestries are far more diverse due to more recent genetic admixture compared to individuals with East Asian ancestries [3]. This means that individuals who identify as Latin American or African American may have genomes composed of linkage disequilibrium (LD) blocks from Africa, Europe, and the Americas. Admixture and difference in population structure can lead to complexities in studying genetic associations within diverse populations. Nonetheless, these populations need to be studied to better understand disease risk across the globe. The lack of representation of non-European populations in GWAS can lead to further health disparities from non-transferable findings across populations. Since allele and haplotype frequencies differ across populations, susceptibility to disease will vary with these frequencies [4]. Elucidating this susceptibility has grown increasingly important since many disease risk prediction models lose performance accuracy across populations [5]. Having accurate models to predict disease risk is especially important for prostate cancer since disease risk is noticeably different across populations.

Prostate Cancer is the most commonly diagnosed cancer and second leading cause of cancer death among African American men; one in seven black men will be diagnosed with prostate cancer in his lifetime compared to one in nine white men [6]. Risk factors for prostate cancer include age, family history of disease, and African ancestries. The genetic component of prostate cancer is made of rare variants with high penetrance and many common variants with lower penetrance [7, 8]. Adding to the complexity of this analysis, prostate cancer is a remarkably heterogeneous phenotype with various molecular and physical classifications [9]. While prostate cancer susceptibility is increased in African Americans, prostate cancer risk is poorly understood in individuals of East Asian ancestries. The age adjusted incidence rate of prostate cancer in East Asian Americans is nearly three times that of native East Asian individuals [10]. Better understanding of how alleles specific to East Asian populations influence prostate cancer risk could reveal new underlying disease biology. Latin American individuals are the least studied of the three populations. The lack of representation could be due to low rates of incidence of prostate cancer in Central and South America; however, Latin American countries are estimated to have the second highest increase in risk for the disease by 2040 [11].

Of the 67 published prostate cancer GWAS in the National Center for Biotechnology Information (NCBI) GWAS Catalog, only 16 studies included individuals of non-European ancestries [12]. These GWAS have identified 8q24 as a region carrying susceptibility loci for prostate cancer. While some fine-mapping of this region has been done in diverse populations, transcriptome-wide association studies (TWAS) of prostate cancer have not been conducted across diverse populations [13, 14]. TWAS serves as a systematic method for integrating expression quantitative trait loci (eQTL) data from GWAS [15-17]. Since TWAS uses gene expression as an intermediate phenotype, it has a functional advantage over GWAS. One of the largest GWAS published up to this point that included over 140,000 individuals of European ancestries was the Prostate Cancer Association Group to investigate Cancer-Associated Alterations in the Genome (PRACTICAL) Consortium [8]. Two of the largest gene-based association studies specific to prostate cancer were TWAS of the PRACTICAL GWAS summary statistics [18, 19]. While these studies provided insight into genes associated with prostate cancer in European subjects, they provided little insight into genes affected by ancestry-specific SNPs in diverse populations. Moreover, broader and more accurate genotype and gene expression imputation panels have been created since the studies in diverse populations were published.

We seek to better understand the genetic architecture of gene expression for prostate cancer in African American, Japanese American, and Latin American populations. To do this, we performed a standard case-control GWAS across 4,769 African American subjects, 2,147 Latin American subjects, and 2,199 Japanese American subjects. We used a deterministic approximation of posteriors for GWAS (DAP-G) to fine-map the 8q24 susceptibility region [20]. In addition to GWAS, we performed TWAS using PrediXcan across 46 tissues from GTEx version 8 and five models from the Multi-Ethnic Study of Atherosclerosis. We replicated our data in an S-PrediXcan application to the PRACTICAL summary statistics [15, 21]. All scripts and notes for this study can be found at https://github.com/WheelerLab/Prostate-Cancer-Study.

Methods

Genotype and phenotype data

Phenotype and genotype data for all individuals in this study were downloaded from the NCBI database of Genotypes and Phenotypes (dbGaP) accession number phs000306.v4.p1. Our project was determined exempt from human subjects federal regulations under exemption number 4 by the Loyola University Chicago Institutional Review Board (project number 2014). Participants were mainly self-reported ethnicities of African Americans, Japanese Americans, or Latin Americans. Cases of cancer within these men were identified by annual linkage to the National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) registries in California and Hawaii. Whole genome genotypying was performed on Illumnia Human 1M-Duov3_B and Human660W-Quad_v1_A array platforms surveying 1,185,051 and 592,839 single nucleotide polymorphisms (SNP), respectively [22, 23]. The final association analyses included 4,769 African American subjects (2,463 cases and 2,306 controls), 2,147 Japanese American subjects (1106 cases and 1093 controls), and 2,199 Latin American subjects (1081 cases and 1066 controls) (Table 1).

Table 1

Population characteristics: Genotype and phenotype data from the three populations went through standard genome-wide quality control and genotype imputation.

Population	African American	Japanese American	Latin American
Pre-QC Individuals	4874	2199	2147
Post-QC Individuals	4769	2199	2147
Cases	2463	1106	1081
Controls	2306	1093	1066
Pre-QC SNPs	1,199,187	657,366	657,366
Post-QC SNPs	1,077,583	540,326	539,366
Post-Imputation SNPs	15,394,464	4,623,264	7,010,834

Quality control and imputation

After we downloaded the data from dbGaP, we divided the subjects into three groups of their self-reported ethnicities. Standard genome-wide quality control was performed separately on each of the three groups using PLINK [24]. In each group, we removed SNPs with genotyping rates < 99%. We then removed SNPs significantly outside of Hardy-Weinberg Equilibrium (P < 1 × 10−6). We also filtered out individuals with excess heterozygosity, removing individuals at least three standard deviations from mean heterozygosity. We then used PLINK to calculate the first ten principal components of each study when merged with three populations from HapMap phase 3 [25]. The first three principal components of each group were used to confirm self-identified ethnicity (S1 Fig).

Following the confirmation of ethnicity, filtered genotypes were imputed using the University of Michigan Imputation Server [26]. All three groups of genotypes were imputed separately using minimac4 and Eagle version 2.3 for phasing. For the Japanese American population, 1000 Genomes Phase 3 version 5 (1000G) was used as a reference panel [27]. Both the African American and Latin American groups were imputed with 1000G and the Consortium on Asthma among African-ancestry Populations in the Americas (CAAPA) [28]. Genotypes for all three populations were filtered for imputation accuracy and minor allele frequency (MAF). SNPs with r2 < 0.8 and MAF < 0.01 were removed from the analysis. The union of genotypes imputed with 1000G and CAAPA in African American and Latin American populations were used for downstream analysis. r2 and MAF filtering took place before the merging of genotypes. For SNPs in the intersection of the two imputation panels, the SNP imputed with CAAPA was selected for the GWAS and PrediXcan analysis due to LD similarity between the CAAPA reference panel and the populations studied. The number of SNPs and individuals in each ethnicity at various steps throughout quality control and imputation are reported in Table 1.

Genome-wide association study

We performed a traditional case-control GWAS of prostate cancer using a logistic regression in PLINK. The first ten principal components were used as covariates to account for global population structure. P < 5 × 10−8 was used as the P-value threshold to denote genome-wide significance. Independent loci were determined and analyzed using deterministic approximation of posteriors for GWAS (DAP-G), an integrative joint enrichment analysis of multiple causal variants [20]. SNPs were clustered into groups with both a cluster posterior inclusion probability (PIP) and an individual SNP PIP. SNPs in the same cluster were identified as linked, and each group of SNPs was considered a locus independent of others. Pairwise LD calculations of r2 were calculated in PLINK [24].

PrediXcan and S-PrediXcan

We used the gene expression imputation method, PrediXcan, to predict genetically regulated levels of expression for each individual across the three populations [15]. We predicted expression using five models built from monocyte transcriptomes of self-identified European (CAU), Hispanic (HIS), African American (AFA) individuals in the Multi-Ethnic Study of Atherosclerosis (MESA). The two other MESA models were built from combined African American-Hispanic (AFHI) data, and all three populations (ALL) [29]. Additionally, we also predicted expression using the 46 multivariate adaptive shrinkage (mashr) models built from 46 tissues in the Gene-Tissue Expression Project (GTEx) version 8 [30, 31]. Ovary, uterus, and vagina were excluded from the total tissues. In the GTEx version 8 prediction models, only tissues from individuals with European ancestries were used. All gene expression prediction models may be found at http://predictdb.org/. We tested the predicted expression levels for association with the case-control status of individuals in each ethnic population using the first ten principal components as covariates. Significant gene-tissue association were determined after performing an adaptive shrinkage using the R package ashr to account for multiple testing [32]. The adaptive shrinkage calculated a local false sign rate (lfsr) for each test. Gene-tissue pairs with lfsr < 0.05 were considered significant. We chose lfsr over traditional false discovery rate because lfsr accounts for both effect size and standard error for each gene-tissue pair. To confirm significant gene-tissue findings, we used COLOC to investigate colocalization between the GWAS and eQTLs [33]. We followed up these finding using the summary statistics version of PrediXcan, S-PrediXcan [21]. We applied the same GTEx prediction models to the PRACTICAL summary statistics, which included over 20.3M SNPs [8].

Results

Genome-wide association studies

To better characterize the genetic architecture underlying prostate-cancer across non-European populations, we performed a case-control GWAS of prostate cancer across 15M SNPs in nearly 5,000 self-identified African American individuals. Additionally, we performed GWAS across 4.6M SNPs in a nearly 2,200 Japanese American individuals and 7.0M SNPs in 2,147 Latin American individuals. Notably, this GWAS includes imputed SNPs from 1000G and CAAPA, two reference panels not available at the time of the original genotyping and GWAS of this case-control study [22, 23, 27, 34].

In our GWAS of 4,769 self-identified African American individuals, we found 110 SNPs to be significantly associated (P < 5 × 10−8) with prostate cancer. Of these 110 SNPs, 108 of them were located at a previously identified region of chromosome 8 (Fig 1A). Of the SNPs on chromosome 8, rs76595456 was the most significantly associated at P = 1.01 × 10−15. The minor allele (T) of rs76595456 is a SNP found only in individuals of recent African ancestries (Fig 2). Four independent clusters were identified by DAP-G on chromosome 8 at 8q24 (Fig 1C). Of the four clusters, two contained SNPs meeting genome-wide significance, one led by rs76595456 (PIP = 0.942) and the other led by rs72725879 (PIP = 0.994) (S1 Table). rs72725879 was also significant in the Japanese American GWAS (Fig 1B and 1D). Two of 110 SNPs (rs10149068 & rs8017723) were found at a novel locus on chromosome 14. These two SNPs on chromosome 14 are linked (r2 = 0.989) and identified as one locus.

Fig 1

Fine-mapping of the top prostate cancer GWAS signals in African Americans and Japanese Americans.

(A & B) depict a LocusZoom plots of GWAS results from African American and Japanese American populations, respectively [35]. The most significant SNPs in both GWAS were in the same chromosome 8 region. (A) is plotted using 1000G AFR 2014 LD, and (B) is plotted using 1000G ASN 2014 LD. The y-axis is the −log10(P) while the x-axis is location on chromosome 8 measured in megabases (Mb). Color represents the LD r2. (C & D) depict the GWAS −log10(P) compared to DAP-G SNP posterior inclusion probabilities (PIP) for the African American and Japanese American populations, respectively [20]. Each point on the plot represents one SNP in each GWAS. The color of each point represents the independent cluster the SNP was assigned to in its respective population. Grey points represent those that were not clustered by DAP-G.

Fig 2

Global allele frequencies of SNPS significantly associates with prostate cancer.

A depiction of the global minor allele frequencies rs76595456 (A), rs72725879 (B), and rs1456315 (C). (A) rs76595456 represents the most significantly associated SNP in the African American GWAS. The minor allele, T, is found only in populations of recent African ancestries. (B) rs72725879 represents the only SNP to be identified by DAP-G in a cluster in both the African American and Japanese American GWAS. (C) rs1456315 represents the most significantly associated SNP in the Japanese American GWAS. rs1456315 (C) is found in strong LD with rs72725879 (B) (r2 = 0.815) in the Japanese American GWAS population. rs1456315 and rs72725879 are not linked in the African American GWAS population (r2 = 0.448). This figure was adapted from one generated using the Geography of Genetic Variants Browser [36]. SNP position on chromosome 8 is labeled using hg19 coordinates from 1000G.

The GWAS of 2,199 Japanese American individuals identified 123 SNPs as genome-wide significant. Every genome-wide significant SNP was located at the 8q24 region of chromosome 8 (Fig 1B). rs1456315 was the most significantly associated SNP in this study (P = 1.40 × 10−13). We identified six distinct clusters of SNPs with DAP-G (Fig 1D). Two of the six clusters contained SNPs meeting genome-wide significance with PIP > 0.5. rs1456315 had not only the lowest P-value, but also the highest PIP (PIP = 0.990) in the Japanese American GWAS (Fig 2). rs1456315 was found to be marginally significant in the African American GWAS (P = 1.29 × 10−7). rs1456315 and rs72725879 are linked (r2 = 0.815) in the Japanese American GWAS population and have similar allele frequencies across East Asian populations while rs1456315 and rs72725879 are less linked (r2 = 0.448) in the African American GWAS population and have divergent allele frequencies across African populations (Fig 2). r2 values between SNPs were calculated using PLINK in each individual case-control study. D′ values were calculated using the 1000G African reference population (AFR) in LDlink [27, 37]. The GWAS of 7M SNPs in 2,147 Latin American individuals identified no genome-wide significant SNPs. chr13:106685795 was the most associated SNP (P = 4.41 × 10−7), located on chromosome 13 (S2 Fig).

Gene-based association studies

We used the gene expression imputation tool, PrediXcan to predict gene expression from genotypes using models built from transcriptomes of 46 tissues in version 8 of GTEx [17, 30]. We also predicted gene expression using five models built from monocyte transcriptome of diverse populations in MESA [29]. After predicting the genetically regulated level of expression for each gene using each of these models, we tested the predicted gene expression for association with the phenotype status of each subject using the first ten principal components as covariates.

In our application of PrediXcan to the 4,769 African American individuals, we predicted expression of 489,459 gene-tissue pairs. We identified two gene-tissue pairs with lfsr < 0.05 and nine gene-tissue pairs with lfsr < 0.10 across all GTEx version 8 mashr prediction models (Table 2; Fig 3). EBPL was the gene in all nine of the gene-tissue pairs. The two most significantly associated gene-tissue pairs by lfsr were found in cerebellar hemisphere (lfsr = 0.0423) and cervical spinal cord (lfsr = 0.0485). The gene-tissue pair with the lowest P-value was KLK3, a gene encoding for prostate-specific antigen, in Tibial Artery (P = 3.31 × 10−6), but the lfsr was not significant (lfsr = 0.921). No gene-tissue pairs in the MESA prediction models significantly associated with prostate cancer in the African American population (S3 Fig).

Table 2

PrediXcan significant genes significant (lfsr < 0.05) gene-tissue pairs from PrediXcan analysis.

lfsr represents the local false sign rate as calculated using adaptive shrinkage [32]. P(PRACTICAL) represents P-value for the gene-tissue pair in the S-PrediXcan analysis of the PRACTICAL summary statistics. Beta(PRACTICAL) represents the effect direction and size predicted from S-PrediXcan. “NA” means that the gene was not tested in the PRACTICAL summary statistics.

Population	Gene	Tissue	lfsr	P	Beta	P (PRACTICAL)	Beta (PRACTICAL)
African American	EBPL	Brain_Cerebellar_Hemisphere	0.0423	5.25E-05	-0.022	0.3	0.006
African American	EBPL	Brain_Spinal_cord_cervical_c-1	0.0485	4.94E-05	-0.032	0.551	0.005
Japanese American	PLCL1	Adrenal_Gland	0.0448	9.72E-07	0.752	0.419	-0.068
Japanese American	PLCL1	Artery_Aorta	0.0435	9.72E-07	0.738	0.419	-0.067
Japanese American	NKX3-1	Brain_Caudate_basal_ganglia	0.0472	1.05E-05	-0.208	3.61E-25	-0.288
Japanese American	FAM227A	Brain_Nucleus_accumbens_basal_ganglia	0.042	9.13E-06	-0.211	NA	NA
Japanese American	NKX3-1	Brain_Putamen_basal_ganglia	0.0475	1.05E-05	-0.212	1.70E-46	-0.215
Japanese American	FAM227A	Brain_Putamen_basal_ganglia	0.0379	8.21E-06	-0.183	NA	NA
Japanese American	NKX3-1	Brain_Substantia_nigra	0.0209	3.80E-06	-0.238	1.18E-20	-0.288
Japanese American	FAM227A	Esophagus_Mucosa	0.0444	9.80E-06	-0.206	NA	NA
Japanese American	NKX3-1	Heart_Left_Ventricle	0.0469	1.05E-05	-0.202	3.61E-25	-0.28
Japanese American	FAM227A	Heart_Left_Ventricle	0.0434	9.22E-06	-0.222	NA	NA
Japanese American	NKX3-1	Liver	0.0468	1.05E-05	-0.198	3.61E-25	-0.274
Japanese American	PLCL1	Lung	0.0485	9.72E-07	0.797	0.419	-0.072
Japanese American	NKX3-1	Muscle_Skeletal	0.0413	8.02E-06	-0.147	3.61E-25	-0.375
Japanese American	PLCL1	Nerve_Tibial	0.031	1.40E-06	0.44	0.537	-0.033
Japanese American	FAM227A	Pituitary	0.0418	9.09E-06	-0.178	NA	NA
Japanese American	FAM227A	Prostate	0.045	1.00E-05	-0.186	NA	NA
Japanese American	COQ10B	Thyroid	0.0418	2.29E-07	-1.138	0.298	0.13

Fig 3

Prostate cancer PrediXcan results for GTEx predicted genes in African American and Japanese American populations.

(A & B) are Manhattan plots of the PrediXcan results using GTEx version 8 mashr gene expression prediction models for the respective African American and Japanese American populations. Each point represents a gene-tissue test for association with prostate cancer via PrediXcan. The y-axis represents the −log10(P) of the gene-tissue test, and the x-axis plots chromosome number. The size of the dot is inversely proportional to its lfsr.

We used PrediXcan to impute and associate gene expression with prostate cancer across 270,813 gene-tissue pairs in GTEx version 8 from 2,199 Japanese American individuals. We found seventeen gene-tissue pairs to be significantly associated with prostate cancer in this population (Table 2). Of these seventeen gene-tissue pairs, four unique genes were identified: PLCL1, NKX3-1, FAM227A, COQ10B. The most significantly associated gene-tissue pair by P-value was COQ10B in Thyroid (P = 2.28 × 10−7). When we applied PrediXcan to the Latin American study, we predicted expression of 411,366 gene-tissue pairs in GTEx version 8. No genes were significantly associated with prostate cancer by lfsr or P-value (S4 Fig).

We attempted to replicate our PrediXcan findings by applying S-PrediXcan to GWAS summary statistics of the PRACTICAL meta-analysis of over 140,000 individuals of European ancestries performed by Schumacher et al. [8]. 424,518 gene-tissue pairs were tested from the summary statistics using GTEx version 8 prediction models. When we compared these summary level results to our findings, only NKX3-1 replicated with a P-value meeting Bonferroni significance (P < 1.178 × 10−7). Similar to the findings of our TWAS, the direction of effect predicted by S-PrediXcan was negative, associating decreased expression of NKX3-1 with prostate cancer (Table 2). Not enough SNPs were not present in the FAM227A prediction model to reliably predict expression from PRACTICAL summary statistics.

Discussion

Broadly, the findings of this study confirm previously well-established information about the genetics of prostate cancer in diverse populations; however, our study differs from previous ones since we performed the first TWAS of prostate cancer in these populations using mashr prediction models from GTEx version 8 and diverse prediction models from MESA [17, 29]. Two of the main findings of this study were identifying risk loci at chromosome 8q24 and identifying NKX3-1 as a risk gene for prostate cancer. Prostate cancer risk at 8q24 has been well characterized to carry SNPs that are population-specific to African Americans [13, 23, 38]. Previously, up to twelve independent risk signals have been identified at 8q24 in European populations and three significant loci in populations of African ancestries [13, 39] (S2 & S3 Tables).

Our study found four independent clusters of SNPs at this position for the African American study using DAP-G. Two of these four clusters contained SNPs meeting genome-wide significance. Han et al. previously identified rs72725879, rs114798100, and rs111906932 to be three distinct significantly associated loci [13] (S3 Table). All three of these SNPs met genome-wide significance in our study, however, our DAP-G modeling clustered SNPs differently than in the forward selection modeling of Han et al. (Fig 1). In both studies, rs72725879 was identified as an independent risk signal. Han et al. found rs114798100 and rs111906932 to be independent risk loci while DAP-G in our study prioritized rs76595456 as risk signal independent from rs72725879. Interestingly, DAP-G clustered rs114798100 (PIP = 1.21 × 10−5) with rs76595456 (PIP = .942), and rs111906932 was not assigned to any cluster by DAP-G. rs76595456 appears to be frequently co-inherited with rs114798100 (D′ = 0.964 in 1000G AFR) and rs111906932 (D′ = 0.975 in 1000G AFR). The small number of independent signals identified by DAP-G is unsurprising since DAP-G is a more conservative fine-mapping tool that assumes a single causal variant is expected a priori. This difference could also be attributed to sample size since Han et al. had a sample size nearly twice that of our study. Additionally, we found six clusters in the Japanese American study using DAP-G. Of the 102 SNPs assigned to clusters by DAP-G in either African or Japanese American population, only rs72725879 overlapped across populations (Fig 2). rs72725879 has previously been implicated in Asian ancestry-specific risk to prostate cancer [14], and is in high LD (r2 = 0.815) with rs1456315, which was the most significantly associated SNP in the Japanese American GWAS.

With respect to our identification of NKX3-1 in the Japanese American TWAS, NKX3-1 was identified across six tissues including both brain and somatic tissue. NKX3-1 is a well known tumor suppression gene, whose decreased expression has been associated with prostate cancer [40, 41]. In all six tissues, NKX3-1 expression was predicted with a negative direction of effect, associating decreased expression with the phenotype. This direction of effect is replicated both in our S-PrediXcan application to the PRACTICAL summary statistics and previous finding in Japanese populations [42].

rs76595456 is identified as an African ancestry-specific SNP

rs76595456 was the most significantly associated (P = 1.01 × 10−15) SNP in our study. Its minor risk allele is specific to populations of recent African ancestries (MAF = 0.1150), and it is absent in both Asian and European populations [27] (Fig 2). The SNP had an individual PIP of 0.942 and was found in a cluster with seven other SNPs bearing a cluster PIP of 0.995. The SNP is an intron variant of PCAT2, a well-established prostate cancer associated transcript [13].

Novel gene associations implicated by TWAS

In our TWAS of prostate cancer across African Americans, we report EBPL as a gene significantly associated (lfsr = 0.0423) with prostate cancer in this population. EBPL made up the top twenty gene-tissue pairs by lfsr (lfsr ranging from 0.0423-0.153). It was the most associated gene reported across all five MESA gene expression prediction populations by both P-value (P = 4.67 × 10−6 in AFA) and lfsr (lfsr = .106 in AFHI). The highest gene-tissue COLOC P4 value, the probability that the GWAS and eQTL signal are colocalized, was 0.18. In the S-PrediXcan analysis of the PRACTICAL data, the gene did not replicate. This failure to replicate could be attributed to difference genetic architecture across the African American test population and the European PRACTICAL population.

The TWAS of prostate cancer in the Japanese American population revealed four unique genes associated with prostate cancer. The previously discussed NKX3-1 is a well-established tumor suppressor gene whose decreased expression is known to progress the aggressiveness of the tumor [41, 43]. COQ10B has not been associated with prostate cancer previously according to the NCBI GWAS Catalog [12]. PLCL1 has been nominally associated with prostate cancer through a SNP x SNP interaction study [44]. Neither of these genes replicated in the larger S-PrediXcan analysis. FAM227A provides an interesting situation since it has been previously identified to be associated with prostate cancer in a GWAS of prostate cancer in Middle Eastern populations [45]. Additionally, it was the only significant gene-tissue pair to be found in prostate tissue. SNPs were not present in the PRACTICAL summary stats to generate a reliably predicted level of expression of FAM227A using S-PrediXcan. Despite having nearly 15 million more SNPs in the PRACTICAL summary statistics compared to our GWAS and TWAS of Japanese American populations, SNPs were not genotyped or imputed at this location on chromosome 22 in the PRACTICAL study. One of the SNPs in this model, rs16999186, has a divergent allele frequency across Japanese (MAF = 0.114) and European (MAF = 0.0169) populations [27]. Since this frequency in European populations falls near the quality control MAF threshold in the original PRACTICAL study and below the genotyping threshold for the European-specific designed genotyping array, this SNP could easily be missed in larger studies [8]. FAM227A also lies within approximately 150kB of SUN2 (P = 1.18 × 10−5;lfsr = .108) a gene marginally significant in our MESA-HIS prediction model. SUN2 has recently been identified as a gene whose decreased expression has been significantly associated with prostate cancer in Japanese populations [46].

Conclusion

In summary, this study of 4,769 African American and 2,199 Japanese American individuals identifies potential population-specific risk loci for prostate cancer in people of recent African or East Asian ancestries. Since its minor risk allele is found only in populations of recent African ancestries, rs76595456 is suggested as a potentially novel risk SNP to prostate cancer in African Americans. Furthermore, the identification of FAM227A as a potential susceptibility gene for prostate cancer in non-European populations highlights growing need for studies of the genetics of prostate cancer in non-European populations. This study’s principal limitations were sample size and ancestry matching in gene expression prediction models. Sample sizes of under 5,000 African American subjects and nearly 2,200 Japanese American subjects pale in comparison to studies of exclusively European populations that are nearly two orders of magnitude larger. Considering that African American men are nearly twice as likely to die from prostate cancer as their white counterparts, the need for larger sample sizes of African American subjects cannot be overstated [6]. Regarding the gene expression prediction models used, the GTEx version 8 prediction models are the most comprehensive set of prediction models to date; however, they are built exclusively from the transcriptomes of European ancestries individuals. Where the models provide accuracy and breadth in capturing common eQTLs, they struggle to predict expression from population-specific eQTLs. The MESA prediction models used capture some of the diversity across populations, but they too are limited by sample size (233 African American individuals, 352 Hispanic individuals, and 578 European individuals) and diversity considering there is no model built from transcriptomes of East Asian subjects [29]. To better understand the genetic processes that underlie prostate cancer in diverse populations, more diverse studies are needed.

Quality control PCA against HapMap.

After merging genotypes with those of four reference populations from version three of the HapMap Project, we performed principal component analysis of all three study populations separately. African American (A) and Japanese American (B) genotypes are plotted with three populations from HapMap: Chinese in Beijing and Japanese in Tokyo (ASN), European ancestries in Utah (CEU), and Yoruba people in Ibadan, Nigeria (YRI). The Latin American genotypes are plotted with Chinese in Beijing and Japanese in Tokyo (ASN), European ancestries in Utah (CEU), and indigenous people of North America (NAT).

(TIF)

Click here for additional data file.

Chromosome 13 GWAS & DAP-G results.

Genome-wide association studies identified no genome-wide significant SNPs. (A) depicts a LocusZoom plots of the most associated GWAS results from Native American population on chromosome 13 [35]. (A) is plotted using 1000G AMR 2014 LD. The y-axis is the -log(P) while the x-axis is location on chromosome 13 measured in megabases. Color represents the LD r2. (B) depicts the results of our GWAS in comparison to DAP-G cluster and PIP for the Latin American population [20]. Each point on the plot represents one SNP in our GWAS. The y-axis is -log(P), and the x-axis is the individual SNP PIP as calculated by DAP-G. The color of each point represents the cluster to which DAP-G assigned it.

(TIF)

Click here for additional data file.

MESA PrediXcan Manhattan plots.

(A, B, & C) are Manhattan plots of the gene-based association study using MESA monocyte gene expression prediction models for the respective African American, Japanese American, and Latin American populations. Each point represents a gene-tissue test from PrediXcan. The y-axis represents the −log10(P) of the gene-tissue test, and the x-axis plots chromosome number. The size of the dot is inversely proportional to its lfsr.

(TIF)

Click here for additional data file.

Latin American PrediXcan Manhattan plots.

Manhattan plot of the gene-based association study using GTEx version 8 mashr gene expression prediction models for the Latin American population. Each point represents a gene-tissue test from PrediXcan. The y-axis represents the −log10(P) of the gene-tissue test, and the x-axis plots chromosome number. The size of the dot is inversely proportional to its lfsr.

(TIF)

Click here for additional data file.

DAP-G clustered SNPs.

At chromosome 8q24, DAP-G calculated PIPs for 12,785 SNPs and 3,878 SNPs in the African American and Japanese American populations, respectively. Of these SNPs with a calculated PIP, 223 SNPs (24 SNPs in African Americans and 199 SNPs in Japanese Americans) were placed into a cluster. Of those placed into a cluster, 102 SNPs at chromosome 8q24 met genome-wide significance in either population. Nine SNPs in the African American study and 93 SNPs in the Japanese American study met genome-wide significance and were placed into a cluster. Of these 102 clustered SNPs meeting genome-wide significance, rs72725879 was the only SNP to overlap across populations. This table contains DAP-G results for the 102 SNPs clustered that met genome-wide significance.

(CSV)

Click here for additional data file.

Chromosome 8 SNPs intersecting with NCBI GWAS catalog.

A table that includes all SNPs on chromosome 8 from our African American and Japanese American study that intersect with SNPs from prostate cancer GWAS summary statistics uploaded to the NCBI GWAS Catalog [12]. There are twenty-four published papers with results uploaded to the NCBI GWAS Catalog containing SNPs intersecting with our studies [8, 44, 46–67].

(CSV)

Click here for additional data file.

Comparative results to Han et Al. (2016).

A table comparing the GWAS results from four SNPs on chromosome 8 in our study to the same four SNPs in Han et al. (2016). Two of these SNPs rs76595456 and rs72725879 are prioritized in our study to be independent loci while Han et al. found rs72725879, rs114798100, and rs111906932 as three independent risk loci. BP is the chromosomal location using human genome build 19 [13].

(CSV)

Click here for additional data file.

31 Jul 2020

PONE-D-20-20022

Multi-ethnic transcriptome-wide association study of prostate cancer

PLOS ONE

Dear Dr. Wheeler,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

1.  The rationale for the study needs to made more clear, particularly in light of previously published work.  What does this study add?  This is especially important as the authors are using some data from previous publications.

2. The authors need to do a more comprehensive literature search on what has already been accomplished in this area (GWAS, eQTLs for risk alleles etc.) and place their study in the context of this previous work.  Details can be added to both the introduction and discussion.

3.  Address reviewer 1's comments on the 8q24 SNP.

4.  Address the other points raised by the reviewers.

Please submit your revised manuscript by Sep 14 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Amanda Ewart Toland, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

3.We note that [Figure(s) 2] in your submission contain [map/satellite] images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

1.    You may seek permission from the original copyright holder of Figure(s) [2] to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

2.    If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors utilize available prostate cancer GWAS data to perform GWAS and TWAS of prostate cancer within 3 racial/ethnic groups. The GWAS seems like an academic exercise as the GWAS results for these 3 populations/studies have already been published. A stronger rationale for doing this and re-publishing the results needs to be provided. They also need to acknowledge previous GWAS analyses of these same data as many studies have been published. The results are also not properly framed with regards to known risk variants or regions. For example, the top SNP in the study at 8q24, rs7659456, is correlated with much stronger signals in African Americans which have been published. It is odd that these other stronger SNPs are not found which suggests issues with the imputation. While the TWAS analysis in these studies is novel, the results aren’t particularly interesting as what was found are essentially known eQTL signals with known risk genes.

Minor points:

In abstract and elsewhere, this is not a cohort but rather multiple case-control studies.

The location of the 8q24 SNP is incorrect in Fig 2

Reviewer #2: The study by Fiorica and colleagues is a careful analsyes of genetic susceptibility to prostate cancer across races. This is an impactful area of research and one that is being addressed by various groups. The current findings will therefore resonate with multiple efforts to define the genetic drivers of prostate cancer.

Perhaps, there are technical improvements that could be made (for example, in terms of ancestry inference) but these are minor concerns and the authors justify their approaches within the confines of the study.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Moray J Campbell

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

27 Aug 2020

We thank the editor and reviewers for the thorough examination of our manuscript and for providing positive and helpful feedback. We appreciate the opportunity to address reviewer comments here. Our responses are prefaced by RESPONSE:

Editor’s Comments (Amanda Ewart Toland, Ph.D.):

1. The rationale for the study needs to made more clear, particularly in light of previously published work. What does this study add? This is especially important as the authors are using some data from previous publications.

RESPONSE:

We thank the editor for highlighting this point and bringing to our attention the need to contextualize our study. In our response to Reviewer 1, we note the novelty of our TWAS and the fact that imputation reference panels have been updated since the time of previous GWAS of this dataset. Our introduction and discussion have been updated:

Lines 68-74

“These GWAS have identified 8q24 as a region carrying susceptibility loci for prostate cancer. While some fine-mapping of this region has been done in diverse populations, transcriptome-wide association studies (TWAS) of prostate cancer have not been conducted across diverse populations [13, 14]. TWAS serves as a systematic method for integrating expression quantitative trait loci (eQTL) data from GWAS [15-17]. Since TWAS uses gene expression as an intermediate phenotype, it has a functional advantage over GWAS.”

Lines 246-250

“Broadly, the findings of this study confirm previously well-established information about the genetics of prostate cancer in diverse populations; however, our study differs from previous ones since we performed the first TWAS of prostate cancer in these populations using mashr prediction models from GTEx version 8 and diverse prediction models from MESA [17, 29].”

2. The authors need to do a more comprehensive literature search on what has already been accomplished in this area (GWAS, eQTLs for risk alleles etc.) and place their study in the context of this previous work. Details can be added to both the introduction and discussion.

RESPONSE:

We appreciate this comment from the editor. After performing a fuller literature review, we have contextualized our study primarily with respect to Han et al. (2016) JNCI. Supplemental Table 3 details the four SNPs we explore in our discussion. We discuss the differences in our findings and potential reasons for these differences in our discussion when we state:

Lines 253-269

“Previously, up to twelve independent risk signals have been identified at 8q24 in European populations and three significant loci in populations of African ancestries [13,38] (Supplemental Table 2).

Our study found four independent clusters of SNPs at this position for the African American study using DAP-G. Two of these four clusters contained SNPs meeting genome-wide significance. Han et al. previously identified rs72725879, rs114798100, and rs111906932 to be three distinct significantly associated loci [13](Supplemental Table 3). All three of these SNPs met genome-wide significance in our study, however, DAP-G clustered SNPs differently than their findings (Fig 1) . In both studies, rs72725879 was identified as an independent risk signal. Han et al. found rs114798100 and rs111906932 to be independent risk loci while DAP-G in our study prioritized rs76595456 as risk signal independent from rs72725879. Interestingly, DAP-G clustered rs114798100 (PIP=1.21x10-5) with rs76595456 (PIP=.942), and rs111906932 was not assigned to any cluster by DAP-G. rs76595456 appears to be frequently co-inherited with rs114798100 (D'=0.964 in 1000G AFR) and rs111906932 (D'=0.975 in 1000G AFR). The small number of independent signals identified by DAP-G is unsurprising since DAP-G is a more conservative fine-mapping tool that assumes a single causal variant is expected a priori.”

3. Address reviewer 1's comments on the 8q24 SNP.

RESPONSE:

We have since relabelled Figure 2 appropriately. We further explore literature regarding the 8q24 region in African American prostate cancer in our discussion and address this point in our response to Reviewer 1.

4. Address the other points raised by the reviewers.

RESPONSE:

We appreciate the careful and thorough review of our manuscript by the reviewers. We have outlined our response to reviewers below.

We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

RESPONSE:

No data availability changes have been made. All scripts, notes, and results from this study can be found at https://github.com/WheelerLab/Prostate-Cancer-Study. We have also submitted our GWAS summary statistics to the NCBI GWAS Catalog for our GWAS in all three populations.

We note that [Figure(s) 2] in your submission contain [map/satellite] images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

RESPONSE:

The Geography of Genetic Variants Browser, the source from which Figure 2 was adapted, has been made public by Creative Commons Attribution License (CC BY 4.0).

Joseph H Marcus, John Novembre, Visualizing the geography of genetic variants, Bioinformatics, Volume 33, Issue 4, 15 February 2017, Pages 594–595, https://doi.org/10.1093/bioinformatics/btw643

Comments to the Author

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors utilize available prostate cancer GWAS data to perform GWAS and TWAS of prostate cancer within 3 racial/ethnic groups. The GWAS seems like an academic exercise as the GWAS results for these 3 populations/studies have already been published. A stronger rationale for doing this and re-publishing the results needs to be provided.

RESPONSE:

We appreciate Reviewer 1’s comment to provide a stronger rationale for the study. We have addressed this at:

Lines 66-83

“Of the 67 published prostate cancer GWAS in the National Center for Biotechnology Information (NCBI) GWAS Catalog, only 16 studies included individuals of non-European ancestries [12]. These GWAS have identified 8q24 as a region carrying susceptibility loci for prostate cancer. While some fine-mapping of this region has been done in diverse populations, transcriptome-wide association studies (TWAS) of prostate cancer have not been conducted across diverse populations [13,14]. TWAS serves as a systematic method for integrating expression quantitative trait loci (eQTL) data from GWAS [15-17]. Since TWAS uses gene-expression as an intermediate phenotype, it has a functional advantage over GWAS. One of the largest GWAS published up to this point that included over 140,000 individuals of European ancestries was the Prostate Cancer Association Group to investigate Cancer-Associated Alterations in the Genome (PRACTICAL) Consortium [8]. Two of the largest gene-based association studies specific to prostate cancer were TWAS of the PRACTICAL GWAS summary statistics [18,19]. While these studies provided insight into genes associated with prostate cancer in European subjects, they provided little insight into genes affected by ancestry-specific SNPs in diverse populations. Moreover, broader and more accurate genotype and gene expression imputation panels have been created since the studies in diverse populations were published.”

Lines 85-90

“To do this, we performed a standard case-control GWAS across 4,769 African American subjects, 2,147 Latin American subjects, and 2,199 Japanese American subjects. We used a deterministic approximation of posteriors for GWAS (DAP-G) to fine-map the 8q24 susceptibility region [20]. In addition to GWAS, we performed TWAS using PrediXcan across 46 tissues from GTEx version 8 and five models from the Multi-Ethnic Study of Atherosclerosis. We replicated our data in an S-PrediXcan application to the PRACTICAL summary statistics [15, 21].”

Lines 246-250

“Broadly, the findings of this study confirm previously well-established information about the genetics of prostate cancer in diverse populations; however, our study differs from previous ones since we performed the first TWAS of prostate cancer in these populations using mashr prediction models from GTEx version 8 and diverse prediction models from MESA [17, 29].”

They also need to acknowledge previous GWAS analyses of these same data as many studies have been published.

RESPONSE:

We thank Reviewer 1 for helping us recognize the need to contextualize our results. We have included Supplemental Table 3 that includes results of twenty-six different GWAS included in the NCBI GWAS Catalog that contain SNPs that intersect with SNPs in our studies at chromosome 8. The table includes P-values from both our study and the reference GWAS Catalog study. Additionally, it includes sample sizes and ancestries for each study. It should be noted that Han et al. (2016) JNCI, which we discuss in our next comment, is not included in the NCBI GWAS Catalog nor the citations for this set of genotypes and phenotypes in dbGaP. This explains a potential reason that this study was overlooked in our initial literature review.

The results are also not properly framed with regards to known risk variants or regions. For example, the top SNP in the study at 8q24, y, is correlated with much stronger signals in African Americans which have been published. It is odd that these other stronger SNPs are not found which suggests issues with the imputation.

RESPONSE:

Thank you for this suggestion. We recognize that rs72725879 and rs76595456 have previously been identified with more significance in Han et al. (2016) JNCI (N=9,531). We attribute this difference to the fact that the Han et al. study has nearly double the sample size in both their discovery and replication populations. Below is a table comparing the P-values of the Han et al. study and our African American GWAS.

SNP. P-Value (Han et al.) n=9531. P-Value (Fiorica et al.) n=4769

rs72725879. 1.07E-23. 3.68E-12

rs76595456. 1.75E-32. 1.01E-15

rs114798100. 1.61E-33 3.38E-14

rs111906932. 4.32E-10. 2.46E-09

We have further explained this in the paper at lines 253-276, where we write:

“Previously, up to twelve independent risk signals have been identified at 8q24 in European populations and three significant loci in populations of African ancestries [13, 38] (S2 Table and S3 Table).

Our study found four independent clusters of SNPs at this position for the African American study using DAP-G. Two of these four clusters contained SNPs meeting genome-wide significance. Han et al. previously identified rs72725879, rs114798100, and rs111906932 to be three distinct significantly associated loci [13] (S3 Table). All three of these SNPs met genome-wide significance in our study, however, our DAP-G modeling clustered SNPs differently than in the forward selection modeling of Han et al. (Fig 1). In both studies, rs72725879 was identified as an independent risk signal. Han et al. found rs114798100 and rs111906932 to be independent risk loci while DAP-G in our study prioritized rs76595456 as risk signal independent from rs72725879. Interestingly, DAP-G clustered rs114798100 (PIP=1.21x10-5) with rs76595456 (PIP=.942), and rs111906932 was not assigned to any cluster by DAP-G. rs76595456 appears to be frequently co-inherited with rs114798100 (D'=0.964 in 1000G AFR) and rs111906932 (D'=0.975 in 1000G AFR). The small number of independent signals identified by DAP-G is unsurprising since DAP-G is a more conservative fine-mapping tool that assumes a single causal variant is expected a priori. Additionally, we found six clusters in the Japanese American study using DAP-G. Of the 102 SNPs assigned to clusters by DAP-G in either African or Japanese American population, only rs72725879 overlapped across populations (Fig 2). rs72725879 has previously been implicated in Asian ancestry-specific risk to prostate cancer [14], and is in high LD (r2=0.815) with rs1456315, which was the most significantly associated SNP in the Japanese American GWAS.”

In addition, we now include a new supplemental table (S3 Table) containing overlap between our results, and those of the GWAS catalog. Of the 12 different prostate cancer GWAS we found with intersecting genome-wide significant SNPs, seven of the studies contained SNPs with P-values more significant than those identified in our analysis. We attribute this discrepancy to a combination of weaker statistical power due to lower sample size and more stringent imputation filtering in our study. Below are details further explaining these differences in significance:

Hoffman TJ et al. (2015), which accounts for four of the fifteen SNPs with different P-values used r2 ≥ 0.3 (compared to our r2 ≥ 0.8), so there may have been less accurately imputed SNPs in their analysis, potentially inflating P-values. Moreover, their sample included nearly 39,000 non-Hispanic white subjects.

Cheng I et al. (2012), Takata R et al. (2010), and Takata R et al. (2019) all used r2 ≥ 0.3 for their imputation filtering.

Gudmundsson J et al. (2009) and Gudmundsson J et al. (2007) perform their analyses in samples of over 37,000 and 4,500 European individuals, respectively.

Knipe DW et al. (2014) and Schumacher FR et at. (2011) performed their analyses in exclusively European populations.

While the TWAS analysis in these studies is novel, the results aren’t particularly interesting as what was found are essentially known eQTL signals with known risk genes.

RESPONSE:

We thank the Reviewer for recognizing the novelty of our TWAS analysis. We acknowledge that NKX3-1 is a well-characterized prostate cancer risk gene regulated by an enhancer at the 8q24 region that replicated in our S-PrediXcan application of the PRACTICAL summary statistics. Besides this gene, the other four genes we discovered (Table 2) are not located on chromosome 8 and have not been implicated as eQTLs in prostate cancer previously. While these genes did not replicate, there is potentially new biology to be explored in this area in non-European populations. The two largest prostate cancer TWAS to date, Mancuso et al. (2018) Nat Com and Wu et al. (2019) Can Res, use GTEx version 6 prediction models. We note that while these findings might be false positives, we used prediction models that are both more diverse and more novel. We note this at lines 246-250, where we state:

“Broadly, the findings of this study confirm previously well-established information about the genetics of prostate cancer in diverse populations; however, our study differs from previous ones since we performed the first TWAS of prostate cancer in these populations using mashr prediction models from GTEx version 8 and diverse prediction models from MESA [13, 38].”

Minor points:

In abstract and elsewhere, this is not a cohort but rather multiple case-control studies.

RESPONSE:

We appreciate the Reviewer bringing this point to our attention. We have changed the language surrounding each case-control study to reflect this.

The location of the 8q24 SNP is incorrect in Fig 2

RESPONSE:

We apologize for this mistake. The SNP in Figure 2A was mislabeled as rs7659456 when the SNP in our study was rs76595456. The position for rs76595456 was correctly labeled, but there was a missing number in the SNP ID. We have corrected any mislabeled SNPs and changed the language in the Figure 2 caption to read:

Reviewer #2: The study by Fiorica and colleagues is a careful analsyes of genetic susceptibility to prostate cancer across races. This is an impactful area of research and one that is being addressed by various groups. The current findings will therefore resonate with multiple efforts to define the genetic drivers of prostate cancer.

Perhaps, there are technical improvements that could be made (for example, in terms of ancestry inference) but these are minor concerns and the authors justify their approaches within the confines of the study.

RESPONSE: We thank Reviewer 2 for his careful review and appreciation of our work. We acknowledge his comment regarding technical improvements with respect for ancestry inference. We agree that these would be excellent improvements to pursue in future studies with larger sample sizes.

Submitted filename: Fiorica_PONE_Response_to_Reviewers.pdf

Click here for additional data file.

11 Sep 2020

Multi-ethnic transcriptome-wide association study of prostate cancer

PONE-D-20-20022R1

Dear Dr. Wheeler,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Amanda Ewart Toland, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: (No Response)

Reviewer #2: thoughtful comments to generally positive reviews. The comments by the reviewers highlighted certain issues which have been answered

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Moray J Campbell

17 Sep 2020

PONE-D-20-20022R1

Multi-ethnic transcriptome-wide association study of prostate cancer

Dear Dr. Wheeler:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Amanda Ewart Toland

Academic Editor

PLOS ONE