Study Population
In total, 664 patients with histologically confirmed breast cancer were recruited from the Instituto de Cancerología and Hospital General San Juan de Dios, both in Guatemala City. Most of the patients self-identified as “Mestizo”, meaning non-indigenous people of mixed European-Amerindian ancestry who primarily speak Spanish [22]. Overall, the median age at diagnosis of the breast cancer cases was 49 (IQR: 41–61]) (Table 1). The median number of children per study participant was 3 [IQR: 2–4] and the median age at menarche was 13 [IQR; 12–14]. Furthermore, fifty-seven percent of patients were post-menopausal at the time of diagnosis. Of the patients with available family history information, 17% had at least one first- or second-degree relative with breast cancer. The genetic ancestry of patients was determined using genotypes from a SNP microarray and compared to populations of European (EUR), Asian (ASN), or African (AFR) ancestry. Data from a separate population of Guatemalans, self-identified as indigenous, indicated ASN scores of 0.6 to be correlated with high indigenous ancestry (Supplemental Table 1). In our study, ancestry data was available for 575 of the 664 patients. Of these 575 patients, 4.2% (24/575) had an ASN score of < 0.2, 48% (273/575) had an ASN score between 0.2–0.4, 33% (191/575) between 0.4–0.6, and 15% (87/575) had an ASN score > 0.6. We chose to use ASN = 0.5 as a general cut off score for comparing patients with more or less indigenous ancestry.
Table 1
Association of study population characteristics with pathogenic mutations
|
Total, n = 664
|
|
Pathogenic, n = 73
|
Other, n = 591
|
|
|
Median (IQR)
|
Range
|
Median (IQR)
|
Range
|
Median (IQR)
|
Range
|
p-value
|
Age at diagnosis
|
49 (41–61)
|
19–93
|
43 (37–52)
|
23–86
|
50 (41–61)
|
19–93
|
< 0.0001
|
Age at first pregnancy
|
21 (18–25)
|
13–45
|
20 (17–25)
|
15–35
|
21 (18–25)
|
13–45
|
0.40
|
Age at menarche
|
13 (12–14)
|
5–24
|
13 (12–14)
|
9–18
|
13 (12–14)
|
5–24
|
0.33
|
Number of children
|
3 (2–4)
|
0–19
|
3 (2–4)
|
0–10
|
3 (2–5)
|
0–19
|
0.24
|
Number of pregnancies
|
3 (2–5)
|
0–19
|
3 (2–4)
|
0–10
|
3 (2–5)
|
0–19
|
0.11
|
Number of miscarriages
|
0 (0–0)
|
0–5
|
0 (0–0)
|
0–1
|
0 (0–0)
|
0–5
|
0.050
|
Population characteristics are compared between subjects with pathogenic mutations in High and Moderate penetrance genes and those with benign or no variants (Other). The median, interquartile range (IQR) are shown along with p-values for the comparison (Wilcoxon Two-Sample Test).
Germline mutations in known breast cancer susceptibility genes
Targeted sequencing was performed on two different panels of 275 and 468 cancer susceptibility genes, including those known to cause inherited breast and ovarian cancer (BRCA1, BRCA2, PALB2, PTEN, TP53, ATM, BARD1, BRIP1, CHEK2, MSH6, RAD51D, STK11) (Supplemental Table 1). A description of the genes specific to breast cancer along with their relative risk [23] can be found in Table 2. This study examines breast cancer susceptibility genes in a Guatemalan hospital case series of women referred for a tumor biopsy.
We identified 73 pathogenic variants in ATM, BARD1, BRCA1, BRCA2, CHEK2, MSH6, PALB2, and TP53, of which 45 are unique. In addition to mutations in these genes known to have high or medium penetrance in breast and ovarian cancer, we identified 9 rare pathogenic variants in the low/unknown-penetrance genes AXIN2, FH, MLH1, MUTYH, NF1, and SDHB. Mutations in BRCA1 accounted for 50.7% of the non-rare pathogenic variants (37/73), followed by BRCA2 at 20.6% (15/73), PALB2 and TP53 at 6.9% (5/73) each, ATM at 5.5% (4/73), and BARD1 and CHEK2 at 4.1% (3/73) each (Fig. 1A, B).
Association of mutations with age at diagnosis and family history
Of the 664 patients, 538 have “benign” variants (variants confirmed to be non-pathogenic), 82 have pathogenic variants, and 44 have solely variants of uncertain significance (Fig. 1C). Nine patients had both a pathogenic variant and variant of uncertain significance; these patients were counted in just the “pathogenic” group. The associations of study population characteristics with pathogenic mutation are observed in Table 2.
Table 2
Sequenced Breast Cancer Risk Genes and their Inclusion Criteria
Category
|
Gene
|
OR
|
NCCN Risk Assessment for Breast Cancer
|
Inclusion criterion
|
Reference
|
High Risk
|
BRCA1
|
16 (10-28)b
|
Increased risk BRCA
|
|
Momozawa
|
|
BRCA2
|
33 (14-104)b
|
|
|
Momozawa
|
|
PALB2
|
7.5 (5.1-11)a
|
Increased risk of BRCA
|
|
Couch
|
|
PTEN
|
18 (3-750)b
|
Increased risk BRCA
|
|
Momozawa
|
|
TP53
|
8.5 (2.4-46)b
|
|
|
Momozawa
|
Moderate Risk
|
ATM
|
2.8 (2.2-3.6)a
|
Increased risk BRCA
|
|
Couch
|
|
BARD1
|
2.2 (1.3-3.6)a
|
Potentially increased risk BRCA
|
|
Couch
|
|
BRIP1
|
1.6 (1.1-2.4)a
|
Potentially increased risk BRCA
|
|
Couch
|
|
CHEK2
|
2.3 (1.9-2.7)a
|
Increased risk BRCA
|
|
Couch
|
|
MSH6
|
1.9 (1.2-3.3)a
|
Unknown or insufficient risk BRCA
|
|
Couch
|
|
RAD51D
|
3.1 (1.2-7.9)a
|
|
|
Couch
|
|
STK11
|
2.0-4.0c
|
Increased risk of BRCA
|
|
Kurian
|
Low/Moderate
|
NBN
|
1.1 (0.7-1.8)
|
Risk only for 657del%
|
|
Couch
|
|
PMS2
|
0.8 (0.4-1.5)a
|
Unknown or insufficient risk BRCA
|
|
Couch
|
|
RAD51C
|
0.8 (0.5-1.4)a
|
Potentially increased risk of TN BRCA
|
|
Couch
|
|
CDH1
|
5.9-7.3c*
|
Increased risk lobular BRCA
|
|
Kurian
|
Low Risk
|
CDKN2A
|
2.5 (0.8-8.2)a
|
No increased risk for BRCA
|
|
Couch
|
|
MLH1
|
1.2 (0.3-4.2)a
|
Unknown or insufficient risk BRCA
|
|
Couch
|
|
MSH2
|
2.5 (0.8-6.9)a
|
Unknown or insufficient risk BRCA
|
|
Couch
|
|
NF1
|
0.9 (0.6-1.6)a
|
INC risk of BRCA if has clinical NF diagnosis
|
|
Couch
|
|
RAD50
|
0.8 (0.5-1.6)a
|
|
|
Couch
|
Unknown Risk
|
APC
|
|
|
Familial adenomatous polyposis gene
|
|
|
AR
|
|
|
Hormone receptor
|
|
|
AXIN2
|
|
|
Colon cancer risk gene
|
|
|
CDKN1B
|
|
|
Multiple endocrine neoplasia gene
|
|
|
DICER1
|
|
|
DICER1 syndrome gene
|
|
|
EPCAM
|
|
Unknown or insufficient risk BRCA
|
Lynch syndrome gene
|
|
|
ERCC2
|
|
|
DNA repair gene
|
|
|
FH
|
|
|
Kidney cancer gene
|
|
|
HRAS
|
|
|
Costello syndrome gene
|
|
|
MET
|
|
|
Hereditary papillary renal cell carcinoma gene
|
|
|
PMS1
|
|
|
Mismatch DNA repair gene
|
|
|
PTCH1
|
|
|
Basal cell nevus syndrome gene
|
|
|
RB1
|
|
|
Retinoblastoma gene
|
|
|
RET
|
|
|
Multiple endocrine neoplasia 2 gene
|
|
|
SMARCB1
|
|
|
Rhabdoid tumor predisposition syndrome gene
|
|
|
TSC1
|
|
|
Tuberous sclerosis type I gene
|
|
|
TSC2
|
|
|
Tuberous sclerosis type II gene
|
|
|
VHL
|
|
|
Von-Hippel Lindau syndrome gene
|
|
High risk gene are defined as having odds ratios (OR) for association with breast greater than 5; Moderate risk genes have an OR greater than 1; and Low and Low/Moderate genes not been documented to be greater than 1 [23, 32, 34]. *, CDH1 mutations cause the Hereditary Diffuse Gastric Cancer and Lobular Breast Cancer syndrome.BRCA, breast cancer.
|
Clinical characteristics of patients with pathogenic mutations in high or medium penetrance genes are compared to the remainder, that is, patients with benign variants, variants of uncertain significance, and pathogenic variants in low-penetrance genes. Patients with pathogenic mutations in high or medium penetrance genes had a statistically significant earlier age at diagnosis (median, 43 years; IQR: 36.5–51.5) compared to all other patients (median, 49 years; IQR: 41–61) (p < 0.01) (Table 1, Fig. 2A). In addition, 33% of patients with pathogenic mutations in high or medium penetrance genes reported having a family member with breast cancer whereas just 15% of all other patients reported the same (Fig. 2B) (p < 0.001). We found that patients with non-pathogenic mutations were more likely to report having had menopause before diagnosis than patients with pathogenic mutations and also found that patients with pathogenic mutations had fewer miscarriages (Fig. 2A). The complete clinical and pathological characteristics of the 73 pathogenic variants in high and medium penetrance genes are shown in Supplemental Table 3. Although factors such as the number of pregnancies and breastfeeding have been reported to affect breast cancer risk [24], we found no significant difference in these variables between the patients with benign mutations and patients those with pathogenic mutations.
The collective data of all women in our study suggests a significant difference in average age at diagnosis between Guatemalan and US women. The average age at diagnosis for the 664 Guatemalan women is 50.7 years whereas the average for US women is 62 [25].
Recurrent Mutations
A recurrent mutation in this study is defined as any mutation occurring in more than one patient. Eight recurrent pathogenic mutations were discovered in this Guatemala population, including three in BRCA1 (c.799delT, c.212 + 1G > A, c.5123C > A), two in BRCA2 (c.8363G > A, c.2414delC), one in CHEK2 (c.546C > A), one in PALB2 (c.3426_3429del), and one in low-penetrance gene MUTYH (c.1218_1219dup) (Fig. 1B). The BRCA1 variants c.212 + 1G > A (rs80358042) and c.799delT (rs80357724) occurred most frequently, accounting for 19% and 12% of all pathogenic mutations, respectively (Fig. 1B). The c.212 + 1G > A (rs80358042) is a splice site mutation that results in a truncated exon whereas c.798_799del (rs80357724) is a frameshift variant, therefore both mutations result in premature stop codons.
Given the high number of women with the BRCA1 c.212 + 1G > A and c.798_799del, haplotype analysis was conducted to assess the ancestral origin of these mutations (Fig. 3A, B). Genome-wide SNP array data was available to determine the haplotype around a 62 Kb region in 14 of the BRCA1 c.212 + 1G > A carriers and around a 56 Kb region in 9 of the BRCA1 c.798_799del carriers. The BRCA1 c.212 + 1G > A mutation is present on a haplotype common to both Hispanic and European populations so no further determination can be made. However, haplotype analysis revealed a likely European origin for the mutation BRCA1 c.798_799del as the mutation is linked to the C allele of a SNP (rs1799950), present in all 9 carriers, that is much more common in European than in Amerindian or African populations. There was not a high enough number of patients nor informative SNPs to perform a similar analysis on the other recurrent mutations.
Among patients with pathogenic mutations, no significant differences in clinical presentation could be detected between those with a recurrent and a non-recurrent pathogenic mutation. However, an examination of individual mutations revealed that two of the recurrent variants have even earlier ages at diagnosis compared to the average ages at diagnosis for patients with pathogenic mutations. Thus, the average age at diagnosis for patients with BRCA1 c.212 + 1G > A and PALB2 c.3426_3429del mutations were 44 and 39 (Supplemental Table 4).
Variants of uncertain significance
Fifty-four variants were identified in high penetrance breast cancer genes that had no definitive interpretation in ClinVar or Varsome. These 54 variants of uncertain significance (VUS) were seen in 53 patients, with one patient containing 2 VUS and nine patients also carrying pathogenic variants, for a total of 44 (6.6%) unique patients with a VUS only. Thirteen of the variants were found in BRCA1, thirty in BRCA2, seven in PALB2, and two each in PTEN and TP53. Of these 54 VUS, 6 were recurrent mutations and 45 were unique. In ascending order, the percentage of unique variants in each high penetrance gene that is a VUS are 29% (2/7) in TP53, 43% (12/28) in BRCA1, 63% (5/8) in PALB2, 65% (24/37) in BRCA2, and 100% (2/2) in PTEN (Supplemental Table 5).
In addition to these variants of uncertain significance, we have identified pathogenic variants in genes proposed but not demonstrated to be involved in inherited breast cancer. These genes include AXIN2, FH, MLH1, MSH2, MUTYH, NF1, and SDHB. We did not include these variants in our clinical analysis of patients with pathogenic breast cancer variants, as it is unlikely that they are directly linked to breast cancer. We also observed some rare, non-coding variants in high penetrance genes (BRCA1, BRCA2, PALB2, TP53). None of these variants appeared to be pathogenic from our additional risk assessment using Align-GVGD. However, it is possible that these variants may affect splicing or protein function. A list of all variants with their classifications can be found in Supplemental Table 3.
Mammogram use in relation to socioeconomic status
The American Cancer Society measures breast cancer screening rate by the percentage of women 40 and older who had a mammogram in the past 2 years [14]. Of the 445 women over 40 with mammography data, 370 (83%) indicated that they have received a mammogram but only 159 (36%) have had regular screening (Fig. 4A). To provide more context for the rates of mammogram use, we sought to determine whether there is a connection between mammogram use and socioeconomic status. Based on a study in 2015, 80% of the Guatemalan population is estimated to work in the informal sector [26], and, therefore, we used cookstove type as a measure of SES (methods). The relation between the presence of wood-burning stoves in the house and low mammogram usage among women over 40 was extremely significant, with a p-value of 0.0004 (Fig. 4B). Women with wood burning stoves in their houses were less likely to have received prior mammogram screening. We also found a significant relationship between the presence of wood-burning stoves in the house and indigenous ancestry (p = 0.0002), indicating that women with greater genetic indigenous ancestry (ASN score of 0.5 or above) are more likely to also have a wood-burning stove at home (Fig. 4B).
NCCN Guidelines for Genetic Testing
The National Comprehensive Cancer Network (NCCN) guidelines are widely recognized and used as the standard for clinical decision making. The guidelines for breast cancer screening and diagnosis take into account risk factors such as age and family history to recommend genetic testing based on each patient’s likelihood of carrying a pathogenic mutation. We thus sought to determine the utility of the NCCN guidelines in identifying Guatemalan patients with pathogenic mutations. We found that 79% of patients with pathogenic mutations in high or medium penetrance breast cancer genes met the NCCN criteria for genetic screening compared to 61% of all other patients (Table 2, Fig. 2B). Therefore, these guidelines will be useful for clinical genetic decision making.