DNA methylation profiling to assess pathogenicity of BRCA1 unclassified variants in breast cancer

Figures & data

Figure 1. Study Design. (A) Representation of the exonic position of the known mutations in the BRCA1 gene used in this study, including pathogenic (missense and truncating) and test variants. Some variants are represented by multiple tumors. (B) Flow diagram illustrating study design. Rectangular boxes indicate samples, hexagonal boxes represent experimental processes, and diamond-shaped boxes denote bioinformatics processes or analysis.

Figure 2. Methylation array (450K) analysis defines pathogenic and neutral variants. (A) Venn diagram representing the comparison of significant probe lists (FDR <0.05), showing 23 probes only significant in the mutation status analysis. (B) Consensus clustering using 18 probes with a difference in methylation between BRCA1 mutated and BRCAx greater than 5% identified two main clusters, as shown by kmeans plot. (C) The resulting correlation matrix from these two clusters shows that these clusters correlate with mutation status. Unclassified variants in green clustered with either the BRCA1 tumors or the BRCAx tumors.

Table 1. Candidate gene validation by pyrosequencing.

		Array data (β values)							Samples Passed QC				Pyrosequencing Validation (% methylation)
Region Name	Probe ID	Wilcox mut q value FDR^*	wilcox ER q value FDR^#	Wilcox grade q value FDR^$	BRCA1 mutated Median β value	BRCAx Median β value	Methylation difference (β value) [Bx-BRCA1]	SNP in probe targeted CpG?	Total (/96)	BRCA1 (/28)	BRCAx (/47)	Test set (/21)	Wilcox BRCA1 mutated vs. BRCAx P value	Median methylation BRCA1 mutated	Median methylation BRCAx	Methylation difference (%) [Bx-BRCA1]	ICC: array β values vs. Pyrosequencing %^†
BACH2	cg24667115	0.0494	0.0771	0.2378	0.28	0.51	0.23		64	19	27	18	0.0070	35.4	51	15.6	0.717
C8orf31	cg03029255	0.0462	0.1118	0.1711	0.71	0.41	-0.3		47	16	11	20	0.0015	73.6	26.2	-47.4	0.715
C17orf108	cg02502358	0.0494	0.0571	0.4768	0.16	0.07	-0.09		59	19	29	11	0.0454	49.6	10.4	-39.2	0.666
LOC654342	cg21645762	0.0494	0.0542	0.2759	0.77	0.61	-0.16		34	9	18	7	0.00003	88.1	73.8	-14.3	0.642
ch13q34	cg12472473	0.0350	0.7011	0.5566	0.91	0.75	-0.16	Y	53	19	25	9	0.0428	66.1	31.9	-34.2	0.477
chr1p34.2	cg07405182	0.0462	0.2613	0.2437	0.87	0.59	-0.28		48	11	27	10	0.1791	45	62	17	0.469
chr15q22.31	cg08036487	0.0462	0.0542	0.2759	0.75	0.56	-0.19		42	11	25	6	1.0000	52.5	51	-1.5	0.431
chr3q25.2	cg02071853	0.0462	0.0633	0.5381	0.72	0.51	-0.21		36	10	19	7	0.1644	84	71.2	-12.8	0.360
ZNF319	cg06866605	0.0494	0.0771	0.2696	0.9	0.94	0.04		53	12	31	10	0.2972	81.1	83.8	2.7	0.260
chr6q27	cg26074411	0.0494	0.1382	0.1243	0.9	0.85	-0.05		33	6	19	8	0.3319	98.8	88.6	-10.2	0.236
INTS1	cg11644627	0.0490	0.0810	0.8266	0.81	0.89	0.08		52	15	28	9	0.9498	76.4	71.6	-4.8	0.111
chr8p23.1	cg04039177	0.0228	0.2284	0.7353	0.57	0.41	-0.16	Y	45	12	25	8	0.4134	74	68	-6	0.164
DLC1	cg00933411	0.0494	0.0849	0.7508	0.37	0.23	-0.14		33	9	20	4	0.1272	46.6	28.3	-18.3	0.119
TJAP1	cg16759204	0.0358	0.1441	0.1885	0.7	0.57	-0.13		51	13	29	9	0.5013	80.6	87.9	7.3	0.111
PLAGL1	cg02697107	0.0494	0.6577	1.0000	0.86	0.81	-0.05		52	15	28	9	0.7529	88.1	84.1	-4.0	0.103
CRTAC1	cg23801028	0.0437	0.1168	0.4680	0.47	0.63	0.16		66	19	34	13	0.6278	71.7	65.9	-5.8	0.033
chr16q24.1	cg01385438	0.0448	0.0602	0.3067	0.9	0.83	-0.07		X	x	x	x	ND	ND	ND	ND	ND
KRT6C	cg09440385	0.0494	0.0734	0.5752	0.96	0.91	-0.05		X	x	x	x	ND	ND	ND	ND	ND

* False Discovery Rate corrected P value for Wilcoxon rank sum test: BRCA1 mutated tumors vs. BRCAX tumors, β values, array samples

^# False Discovery Rate corrected P value for Wilcoxon rank sum test: ER positive tumors vs. ER negative tumors, β values, array samples

^$ False Discovery Rate corrected P value for Wilcoxon rank sum test: grade 1+2 vs. grade 3, β values, array samples

^† Intraclass Correlation Coefficient (ICC) calculated using array β values and pyrosequencing methylation percentages for the same sample (n between 26 and 54, dependent on assay)

ND Not Done - assay design was not possible/could not be optimized

Figure 3. Validation of LYRM9, BACH2, LOC654342, and C8orf31 loci. (A) Strip charts show array β values for the four validated loci plotted by mutation status. Blue dotted line is plotted at the median of BRCA1 and BRCAx samples. ER negative samples are colored red, ER positive samples are colored blue, and those samples for which ER status is unknown are colored green [BRCA1 (n=18), BRCAx (n=19), BRCA1 test variant (n=17), except BACH2, where BRCA1 n=17]. (B) Strip charts showing the pyrosequencing methylation value of the independent group of samples validating the difference observed on the array [LYRM9:BRCA1 (n=19), BRCAx (n=29), test variant (UV) (n=11); BACH2:BRCA1 (n=19), BRCAx (n=27), test variant (UV) (n=18); C8orf31: BRCA1 (n=16), BRCAx (n=11), test variant (UV)(n=20). LOC654342:BRCA1 (n=9), BRCAx (n=18), test variant (UV) (n=7)].

Table 2. Logistic regression analysis of candidate genes.

Assay	Logistic Regression P values						Number of samples included in logistic regression^*
	mut~meth	mut~meth^+ER		mut~meth^+ER+grade
		Meth	ER	meth	ER	grade
BACH2	0.0265	0.0607	0.0002	0.0848	0.0131	0.0976	61
C8orf31	0.0003	0.0063	0.0101	0.0131	0.0866	0.3721	48
C17orf108	0.0166	0.1010	0.0002	0.1145	0.0495	0.0146	49
LOC654342	0.0038	0.0301	0.0043	0.0829	0.0645	0.0516	37

* only samples with complete data (methylation, ER status, and grade) were used in these logistic regressions

Figure 4. Combined methylation likelihood ratios. LRs were calculated for each sample using the model including only methylation. The log of the combined LR is plotted here corresponding to each variant assayed.

Table 3. Combined summary predictions of all unknown variants analyzed.

Number of tumors	Number of independent tumors^*	HGVS Nuc	HGVS Prot	Prior probability	Current posterior probability^†	Current Classification^†	Posterior probability using methylation only	Classification using methylation only
3	3	c.1486C>T	p.Arg496Cys	0.02	0.00089	Class 1	0.00009	Class 1
3	2	c.2521C>T	p.Arg841Trp	0.02	2.29E-12	Class 1	0.0005-0.00006	Class 1
3	3	c.4963T>C	p.Ser1655Pro	0.03	^∞	Class 3	0.39565	Class 3
2	1	c.4103C>T	p.Ala1368Val	0.02	0.00163	Class 2	0.010-0.004	Class 2
2	1	c.823G>A	p.Gly275Ser	0.02	0.00427	Class 2	0.002-0.006	Class 2
2	2	c.5194-12G>A	IVS	0.34	0.99999	Class 5	0.99638	Class 5
2	2	c.4039A>G	p.Arg1347Gly	0.02	2.04E-12	Class 1	0.00967	Class 2
1	1	c.641A>G	p.Asp214Gly	0.02	0.07426	Class 3	0.00065	Class 1
1	1	c.1984_1992del	p.His662-Arg664del	0.02	0.06058	Class 3	0.04922	Class 2
1	1	c.2912A>G	p.His971Arg	0.02	0.00025	Class 1	0.03682	Class 2
1	1	c.203T>G	p.Ile68Arg	0.66	^∞	Class 3	0.01215	Class 2
1	1	c.4185+9C>T	IVS	0.02	0.00232	Class 2	0.00258	Class 2
1	1	c.4484+2_4484+3ins8	IVS	0.97	0.98991	Class 4	0.97881	Class 4
1	1	c.4485-8C>T	IVS	0.04	^∞	Class 3	0.00884	Class 2
1	1	c.5467+5G>C	IVS	0.34	0.19022	Class 3	0.13048	Class 3
1	1	c.593+16C>G	IVS	0.02	^∞	Class 3	0.14272	Class 3
1	1	c.454C>T	p.Leu152Phe	0.02	^∞	Class 3	0.00016	Class 1
1	1	c.4991T>C	p.Leu1664Pro	0.03	1.18E-05	Class 1	0.22327	Class 3
1	1	c.1534C>T	p.Leu512Phe	0.02	0.00014	Class 1	0.28730	Class 3
1	1	c.4955T>A	p.Met1652Lys	0.66	^∞	Class 3	0.96673	Class 4
1	1	c.3708T>G	p.Asn1236Lys	0.02	^∞	Class 3	0.01645	Class 2
1	1	c.1036C>T	p.Pro346Ser	0.02	0.00427	Class 2	0.00046	Class 1
1	1	c.2180C>T	p.Pro727Leu	0.02	0.00017	Class 1	0.00352	Class 2
1	1	c.5096G>A	p.Arg1699Gln	0.66		Intermediate risk - based on segregation analysis (Spurdle et al., 2015)	0.97998	Class 4
1	1	c.5284A>G	p.Arg1762Gly	0.03	^∞	Class 3	0.00247	Class 2
1	1	c.551C>T	p.Ser184Phe	0.02	^∞	Class 3	0.01149	Class 2
1	1	c.1423A>T	p.Ser475Cys	0.02	0.00395	Class 2	0.02111	Class 2

* only tumors from different individuals can be considered independent

† based on segregation analysis alone (Arg1699Gln) or multifactorial likelihood analysis incorporating segregation, pathology, and other data points (see supplementary Table S2)

∞ combined LR does not pass thresholds recommended as per ENIGMA BRCA classification guidelines (http://www.enigmaconsortium.org/), namely LR of <0.5 (to reach final Class 2 or 1), or >2.0 (to reach final Class 4 or 5) and so should be considered Class 3 (uncertain).