Results

Clinical Characteristics and MPT Combinations
460 individuals (106 [23%] males and 354 [77%] females) in 440 families had been diagnosed with 1,143 primary tumors distributed among 87 categories according to site and cell of origin. The most frequent tumor types are illustrated in Table 2 (comprehensive lists are provided in Tables S1 and S5). Representing 24.6% of the total, breast cancer was the most frequent tumor, and colorectal cancer was the second (9.9%). Prior genetic testing is described in Table S6, and reasons for non-detection of the relevant variant are illustrated in Figure 2.
Figure 2 Molecular Investigations Initiated by Clinical Services with Inferred Reasons for Non-detection of Variants
Table 2 Most Frequent Tumors and Tumor Combinations in the Series
Tumor Category Count Percentage (%)
>5% Total (n = 1,143)
Breast 281 24.6
Colorectal 113 9.9
Kidney 83 7.3
NMSC 67 5.9
Ovary 58 5.1
>1% Total (n = 883)
Breast-colorectal 51 5.8
Breast-NMSC 35 4.0
Breast-ovary 34 3.9
Breast-endometrium 33 3.7
Breast-hem lymphoid 26 2.9
Breast-melanoma 24 2.7
Breast-thyroid 23 2.6
Endometrium-ovary 19 2.2
Breast-kidney 18 2.0
Colorectal-NMSC 14 1.6
Breast-lung 12 1.4
NMSC-hem lymphoid 11 1.2
Breast-soft tissue sarcoma 10 1.1
Colorectal-endometrium 9 1.0
Kidney-pituitary 9 1.0
Kidney-thyroid 9 1.0
Melanoma-NMSC 9 1.0
The following abbreviations are used: hem lymphoid, hematological lymphoid; and NMSC, non-melanoma skin cancer (including basal cell carcinoma and squamous cell carcinoma).
The occurrence of any two discordant primary tumors in the same individual was considered a tumor combination, and a total of 883 combinations and 327 combination types were observed (individuals with three or more discordant tumors had multiple combinations). 206 (63%) combination types occurred once, and 53 (16.2%) occurred twice. The 68 (20.8%) combination types occurring three or more times are illustrated in Figure 3. The most frequent combination type was breast and colorectal cancer, which represented 5.8% of the total combinations. All combination types making up ≥1% of the total are shown in Table 2.
Figure 3 Most Frequent Tumor Combination Types
Combination types occurring fewer than three times are not included. Abbreviations are as follows: pheo, pheochromocytoma; GI NET, gastrointestinal neuroendocrine tumor; hem myeloid, hematological myeloid; PNET, pancreatic neuroendocrine tumor; hem lymphoid, hematological lymphoid; and NMSC, non-melanoma skin cancer (including basal cell carcinoma and squamous cell carcinoma).
To compare the distributions of tumor combinations in our MPT study cohort with a population-based dataset, we compared 313 MPT cohort individuals comprising 523 combinations with 471 individuals comprising 574 combinations in the East Anglia Cancer Registry data (Table S7). There was a significant difference (χ2 p value < 0.05) in the frequency of tumor combinations in 6/12 combination types that individually represented >1% of the total MPT cohort. Breast cancer in combination with ovarian, thyroid, lymphoid hematological, or kidney cancer was overrepresented in the MPT cohort, whereas breast cancer in combination with non-melanoma skin was underrepresented.
Information regarding previous genetic testing was available for 405/440 (92%) probands. No molecular investigations had been performed in 91 (20.7%). 159 (36.1%) had undergone BRCA1 and BRCA2 testing, 87 (19.8%) had been assessed for Lynch syndrome (where microsatellite instability [MSI] and/or immunohistochemistry [IHC] analysis is considered an assessment), and 159 (20.7%) had had another germline genetic test. The mean number of genes analyzed (where MSI or IHC is considered an analysis of four Lynch syndrome genes) was four. Samples from 79 (18%) probands had undergone sequencing with a multi-gene panel assay, and the mean number of genes analyzed with these assays was 13.8.

Genetic Findings

SNVs and Indels
Variant filters applied to annotated VCF files produced 89 unique variants in 119 individuals for further ACMG-guideline-based assessment. Of these, 22 (42 occurrences) could be classified as pathogenic, 23 (24 occurrences) could be classified as likely pathogenic, 24 (27 occurrences) could be classified as a variant of uncertain significance (VUS), and 20 (26 occurrences) could be classified as likely benign. Six occurrences of P/LP variants occurred in two members of the same family, and only three of these contributed to the detection rates quoted below. No pathogenic non-coding variants were identified.
Overall, 63 variants in 17 genes in 61 (13.9%) probands were assessed as P/LP (summary in Table 3; full description with phenotype and previous testing in Table S6). Most were nonsense or frameshift variants. Individuals with variants in moderate-risk CPGs CHEK2 (MIM: 604373; n = 14) and ATM (MIM: 607585; n = 10) were the most frequent; one homozygote was detected for CHEK2: c.1100delC (p.Thr367Metfs) (Ensembl: ENST00000328354; GenBank: NM_007194.3]; annotated in our data as c.1229delC [p.Thr410fs] [Ensembl: ENST00000382580; GenBank: NM_001005735.1]). Individuals with variants in BRCA2 (MIM: 600185; n = 6), PALB2 (MIM: 610355; n = 6), FH (MIM: 136850; n = 5), NF1 (MIM: 613113; n = 4), NTHL1 (MIM: 602656; homozygous, n = 3), MAX (MIM: 154950; n = 2), PTEN (MIM: 601728; n = 2), SDHB (MIM: 185470; n = 2), BMPR1A (MIM: 601299; n = 1), BRCA1 (MIM: 113705; n = 1), CDKN1B (MIM: 600778; n = 1), EXT2 (MIM: 608210; n = 1), MLH1 (MIM: 120436; n = 1), MSH2 (MIM: 609309; n = 1), and PMS2 (MIM: 600259; n = 1) were also noted.
Table 3 Summary of Filtered SNVs and Indels Deemed Pathogenic or Likely Pathogenic by ACMG Assessment
Gene RefSeq mRNA ID No. of Occurrences No. of Individuals with Associated Tumor Variant Description Consequence
ATM GenBank: NM_000051 1 1 c.193C>T (p.Gln65∗) stop gain
ATM GenBank: NM_000051 1 1 c.5623C>T (p.Arg1875∗) stop gain
ATM GenBank: NM_000051 1 0 c.6583+1G>A splice site (donor)
ATM GenBank: NM_000051 1 0 c.6866-6867insT (p.Ser2289Serfs) frameshift
ATM GenBank: NM_000051 1 1 c.748C>T (p.Arg250∗) stop gain
ATM GenBank: NM_000051 1 1 c.8147T>C (p.Val2716Ala) missense
ATM GenBank: NM_000051 1 0 c.8405delA (p.Gln2802fs) frameshift
ATM GenBank: NM_000051 1 0 c.5821G>C (p.Val1941Leu) missense
ATM GenBank: NM_000051 1 0 c.8122G>A (p.Asp2708Asn) missense
ATM GenBank: NM_000051 1 1 c.7775C>G (p.Ser2592Cys) missense
BMPR1Aa GenBank: NM_004329 1 1 c.730C>T (p.Arg244∗) stop gain
BRCA1 GenBank: NM_007300 1 1 c.1961−1962insA (p.Lys654fs) frameshift
BRCA2 GenBank: NM_000059 1 0 c.4525C>T (p.Gln1509∗) stop gain
BRCA2 GenBank: NM_000059 1 1 c.5682C>G (p.Tyr1894∗) stop gain
BRCA2 GenBank: NM_000059 1 0 c.6275−6276delTT (p.Leu2092fs) frameshift
BRCA2 GenBank: NM_000059 1 1 c.6402−6406delTAACT (p.Asn2135Leufs) frameshift
BRCA2 GenBank: NM_000059 1 0 c.6535−6536insA (p.Val2179fs) frameshift
BRCA2 GenBank: NM_000059 1 1 c.1805−1806insA (p.Gly602fs) frameshift
CDKN1B GenBank: NM_004064 1 0 c.148−149delAG (p.Arg50fs) frameshift
CHEK2 GenBank: NM_001005735 3 1 c.1392delT (p.Leu464fs) frameshift
CHEK2 GenBank: NM_001005735 10 6 c.1229delC (p.Thr410fs) frameshift
CHEK2 GenBank: NM_001005735 1 1 c.1051+1C>T splice site (donor)
CHEK2 GenBank: NM_001005735 1 0 c.784delG (p.Glu262fs) frameshift
CHEK2 GenBank: NM_001005735 1 1 c.562C>T (p.Arg188Trp) missense
EXT2 GenBank: NM_000401 1 0 c.613C>T (p.Gln205∗) stop gain
FH GenBank: NM_000143 3 0 c.1433−1434insAAA (p.Lys477_Asn478insLys) in-frame insertion
FH GenBank: NM_000143 1 1 c.320A>C (p.Asn107Thr) missense
FHb GenBank: NM_000143 1 0 c.521C>G (p.Pro174Arg) missense
MAX GenBank: NM_002382 1 1 c.289C>T (p.Gln97∗) stop gain
MAXb GenBank: NM_002382 1 1 c.1A>G (p.Met1Val) start loss
MLH1 GenBank: NM_000249, NM_001258273 1 1 c.1884−1G>A splice site (acceptor)
MSH2 GenBank: NM_000251 1 0 c.1452−1455insAATG (p.Leu484-Met485fs) frameshift
NF1 GenBank: NM_001042492 1 1 c.1541−1542delAG (p.Gln514fs) frameshift
NF1 GenBank: NM_001042492 1 1 c.4620delA (p.Ala1540fs) frameshift
NF1 GenBank: NM_001042492 1 1 c.5831delT (p.Leu1944fs) frameshift
NF1 GenBank: NM_001042492 1 1 c.7768-7769insA (p.His2590fs) frameshift
NTHL1c GenBank: NM_002528 3 3 c.268C>T (p.Gln90∗) stop gain
PALB2 GenBank: NM_024675 4 3 c.3113G>A (p.Trp1038∗) stop gain
PALB2 GenBank: NM_024675 1 1 c.3116delA (p.Asn1039fs) frameshift
PALB2 GenBank: NM_024675 1 1 c.62T>G (p.Leu21∗) stop gain
PMS2a GenBank: NM_000535 1 1 c.741−742insTGAAG (p.Pro247_S248fs) frameshift
PTEN GenBank: NM_000314 1 1 c.1003C>T (p.Arg335∗) stop gain
PTEN GenBank: NM_000314 1 1 c.697C>T (p.Arg233∗) stop gain
SDHB GenBank: NM_003000 1 1 c.223+1C>A splice site (donor)
SDHB GenBank: NM_003000 1 1 c.689G>A (p.Arg230His) missense
This list incorporates one individual per family. See Table S6 for more comprehensive description.
a Occurring in the same individual.
b Occurring in the same individual.
c Homozygous.
The 61 P/LP SNV and indels detected by WGS were confirmed by a second analysis (TCP for 51 variants and Sanger sequencing for ten variants).
Pre-testing information was available for 57/63 P/LP variants, 41/57 (71.9%) of which occurred in an individual who had at least one previous genetic test and 7/57 (12.3%) of which were eventually detected by clinical services. No P/LP variants were observed in genes that had previously been tested in a sample from the relevant individual by diagnostic services (Figure 2). The mean number of genes tested in those with a P/LP variant was 5.3, which was not significantly different from that in probands without such variants detected (Student’s t test p = 0.396).
Of the 61 probands identified with a P/LP variant, 36 (59%; 8.2% of all probands) had previously been diagnosed with a tumor typically associated with the relevant CPG. A further eight (1.8%) probands were found to harbor a VUS and had been diagnosed with an associated tumor.
Two probands harbored two P/LP variants in multiple CPGs. One individual with colorectal adenocarcinoma at age 50 years and breast cancer at 57 years carried a PMS2 frameshift variant (c.741−742insTGAAG [p.Pro247_Ser248fs] [Ensembl: ENST00000265849; GenBank: NM_000535.6]) and a BMPR1A nonsense variant (c.730C>T [p.Arg244∗] [Ensembl: ENST00000372037; GenBank: NM_004329.2]). Immunohistochemistry of the bowel tumor showed loss of PMS2; MSI was also demonstrated, leading to diagnostic sequencing of PMS2, although there was no family history of neoplasia other than an ovarian cancer in a second-degree relative after age 70 years. The proband had previously undergone surveillance colonoscopy for inflammatory bowel disease, resulting in the identification of a number of polyps; however, there was no evidence from histology reports that these were juvenile polyps. Additionally, an individual with bilateral pheochromocytoma at ages 16 and 35 years and no reported family history of neoplasia was identified with variants in FH (c.521C>G [p.Pro174Arg] [Ensembl: ENST00000366560; GenBank: NM_000143.3]) and MAX (c.1A>G [p.Met1Val] [Ensembl: ENST00000358664; GenBank: NM_002382.4]).29 The latter variant is predicted to abolish the MAX initiation codon, and previous analysis of tumor tissue from an individual carrying it demonstrated loss of the wild-type allele and a lack of full-length MAX protein product.30

Coverage and Comparison with Panel
Mean depth in WGS data of coding bases in the 83 genes analyzed was 35× (SD = 7.5), and 100% were covered at ≥10×. Coverage was also considered for 68 of the genes also sequenced by the TCP. In WGS data, 100% of target bases were covered at ≥10×, and the mean depth was 35.3 (SD = 7.4). Coverage analysis pertaining to those 68 genes from the 411 (89.3%) participants also undergoing sequencing with the TCP showed 99.1% target bases at ≥10× and a mean depth of 807.3 (SD = 793.2).
A comparison of the variant detection was performed on the basis of the 105 ACMG-assessed SNVs and indels that were detected by WGS and were within a gene sequenced by the TCP. 99/105 variants were called from TCP data with quality indicators sufficient to pass filters used for the WGS data. Five undetected variants—including one P/LP PMS2 variant (c.741−742insTGAAG [p.Pro247_Ser248fs] [Ensembl: ENST00000265849; GenBank: NM_000535.6]), where 58/202 (20.6%) reads contained the insertion—were indels for which IGV review showed a VAF below the threshold for filtering. One undetected variant in TMEM127 (MIM: 613403) (c.665C>T [p.Ala222Val] [Ensembl: ENST00000258439; GenBank: NM_017849.3]) was covered by only two reads.
The filtering and assessment process applied to WGS data was also used for variants called from TCP data generated from the same 411 individuals. 108/110 TCP variants that passed filters and went forward for ACMG assessment were also called from WGS data, meaning that two variants (assessed as pathogenic) were not detected by WGS. This was because the VAF was marginally below the filtering threshold of 33% for ATM (c.2426C>A [p.Ser809∗] [Ensembl: ENST00000278616; GenBank: NM_000051]) (7/22 [32%] reads) and MAX (c.97C>T [p.Arg33∗] [Ensembl: ENST00000358664; GenBank: NM_002382]) (9/29 [31%] reads).

Comparison of MPT WGS SNV and Indel Detection with gnomAD Dataset
In our dataset, 52 truncating or splice-site variants were observed in 440 MPT probands, whereas 298 were observed in 8,992 gnomAD genomes; the latter is based on observed variant frequency estimates adjusted to reflect sex distribution of the MPT series (13.6% for the MPT dataset versus 3.3% for the gnomAD dataset; χ2 = 84.903, p = < 0.0001). 41 truncating or splice-site CPG variants occurred in a proband with at least one tumor type uncharacteristic of the relevant CPG, and the frequency of such variants in these individuals was also compared with that in gnomAD. This was significantly higher in the MPT probands with uncharacteristic tumors than in gnomAD (41/440 [9.3%] versus 298/8,992 [3.3%]; χ2 = 43.642; p ≤ 0.0001).

SVs
SV analysis revealed six potentially pathogenic variants in 440 (1.4%) probands (Table 4), two of whom had previously been diagnosed with tumors typically associated with variants in the relevant gene. An additional two had no associated tumor but a family history of such tumors in a first-degree relative (colorectal cancer at age 56 years for the individual with a SMAD4 translocation and renal cell carcinoma at age 69 for the individual with the TSC1 duplication). One individual with an inversion of PTEN exon 7 had been diagnosed with breast cancer at age 45 years and had a strong family history of this tumor, which had occurred in her sister (age 57 years), mother (age 57 years), and maternal cousin (age 49 years). The proband’s sister had also been diagnosed with a borderline ovarian mucinous tumor and nasal basal cell carcinoma at 46 and 57 years of age, respectively, but WGS did not detect the PTEN inversion in her sample. Another individual had previously been investigated with germline FH sequencing after the diagnosis of multiple cutaneous leiomyomas and a family history of a first-degree relative undergoing a hysterectomy for uterine leiomyomas. SV analysis revealed a whole-gene deletion of FH.
Table 4 Structural Variants Passing Filtering Steps
Gene Chr Predicted Start Predicted End Algorithm(s) Predicted Consequence after IGV Review Phenotype (Age at Diagnosis) Genes Tested by Clinical Services Year Consulted
FLCN 17 1,7136,696 (Manta), 1,7137,867 (Canvas) 17,134,310 (Manta), 17,134,474 (Canvas) Canvas and Manta deletion of exon 2 breast (46 years) and pulmonary lymphangioleiomyomatosis (47 years) information unavailable unknown
PTEN 10 89,719,837 89,713,996 Manta inversion of exon 7 breast (45 years)a BRCA1 and BRCA2 (single gene) unknown
SMAD4 9 and 18 chr9: 127,732,713 chr18: 48,556,624 Manta translocation with breakpoint within untranslated part of exon 1 CNS (42 years) and colorectal (56 years) in mother PMS2, TP53, and MLH1 (single gene) 2011
TSC1 9 135,807,261 135,803,187 Manta duplication of exon 3 testicular (47 years), prostate (64 years), and lung (70 years) BRCA1 and BRCA2 (single-gene Ashkenazi common mutations) 2016
TSC2 16 2,119,769 1,566,500 Manta inversion with breakpoint in introns 16 and 17 small bowel (42 years) and colorectal (43 years) MSH6 (single gene; IHC also revealed MSH6 loss) 2012
FH 1 242,310,908 237,244,834 Canvas full-gene deletion multiple cutaneous leiomyomata (<55 years)a FH (single gene) 2014
The list incorporates one individual per family. All structural variants are heterozygous. The following abbreviations are used: Chr, chromosome; CNS, central nervous system; IHC, immunohistochemistry.
a Tumor characteristically associated with pathogenic variant in the relevant gene.

Combined Variant Detection Rate
After combining SVs passing our filters and ACMG-assessed P/LP SNVs and indels, we observed a P/LP variant in 67 (15.2%) probands tested. 38 probands (8.6% of total) had such a variant and a typically associated tumor. There was no significant difference in P/LP detection rate between probands diagnosed with a rare tumor and those who hadn’t been (27/136 [19.8%] versus 40/304 [13.1%]; χ2 = 3.2628; p = 0.07087). Of the 55/67 probands for whom a family history was available, there was no cancer diagnosis in a first-degree relative younger than 50 years in 23 individuals (61.8%) and younger than 60 years in 34 individuals (61.8%).
Limited numbers of family members participated in the study, preventing large-scale segregation analysis. Of the 69 P/LP variants (including SVs) of interest detected in probands, the relevant locus was sequenced in a family member on seven occasions. The relevant variant was detected in four of seven family members, two of whom had been diagnosed with a typically associated tumor (breast cancer in PALB2 and BRCA1 variants).