De novo, systemic, deleterious amino acid substitutions are common in large cytoskeleton‑related protein coding regions

Introduction

Genetic damage is largely random and therefore tends to affect the larger, functional regions of the human genome more frequently than the smaller regions (1). For example, a systematic study has revealed that cancer fusion genes, on average, are statistically, significantly larger than other human genes (2,3). The large introns of potential cancer fusion genes presumably allow for many different productive recombination opportunities, i.e., many recombinations that would allow for exon juxtaposition and the generation of hybrid proteins. Smaller cancer fusion genes tend to be associated with the rare types of cancer, for example EWS RNA binding protein 1 in Ewing's sarcoma.

Cytoskeleton-related protein coding regions (CPCRs), including extracellular matrix (ECM) proteins, are among the largest coding regions in the human genome and are heavily mutated in various types of cancer (1,4,5). The possibility that these coding regions would also be commonly vulnerable to de novo mutations, or mutations occurring in relatively recent past-generations, was considered, and therefore had not been included in the conventional single nucleotide polymorphism (SNP) databases. In the present study, CPCR single nucleotide variants (SNVs) that appeared in The Cancer Genome Atlas (TCGA) normal blood and breast cancer (BRCA) sample datasets were identified. ‘Non-unique’ SNVs were discovered to be relatively common among a sample of 31 individuals, and SNVs specifically present in single individuals (i.e., private variations) were also common.

Materials and methods

Basic algorithm

The basic approach is indicated in Fig. 1 and the detailed steps are provided in the supporting online material (SOM) file, ‘Stoll 2016 SOM Figure 1, detailed protocol’; all of the SOM files are hosted at http://www.universityseminarassociates.com/Supporting_online_material_for_scholarly_pubs.php). Whole exome sequence (WXS) sample manifest files were downloaded from the Cancer Genomics Hub Browser (browser.cghub.ucsc.edu; approved NIH dbGaP project number 6300) using the following filters: ‘Breast invasive carcinoma’ under ‘By Disease’; ‘Blood Derived Normal’ and ‘Primary Solid Tumor’ under ‘By Sample Type’; ‘WXS’ under ‘By Library Type’; and ‘GRCH37/HG19’ under ‘By Assembly’. Search results were sorted by ‘Barcode’ so that solid tumor samples were immediately followed by the matched blood sample. The manifest files for matched BRCA and normal blood samples were downloaded and placed into a file labeled ‘Blood’ to be consistent with the language of the BreastBlood.sh coding program (Stoll 2016 SOM Figure 1A-BreastBlood) written specifically for the current study. The manifest files were subsequently processed with the BreastBlood.sh to create .csv output files (from the original .xml manifest files). The original programming code, BreastBlood.sh, and output files (Stoll 2016 SOM Figure 1E-BreastBlood Variants) are provided in the SOM. Note that the output files have been further processed using various macros, also provided in the SOM, to eliminate previously identified variants that will not be included in analysis. BreastBlood.sh was designed to identify variants in cytoskeletal protein-related coding regions (CPCRs). The CPCR list was generated from the study by Parry and Blanck (5), representing the most commonly mutated CPCRs in five cancer datasets that were investigated plus two additional CPCRs that are commonly mutated in the melanoma (skin cutaneous melanoma) dataset, which were not evaluated in the previous study (5). The CPCRs are presented in Table I.

Figure 1. Flow chart representing the processing steps for generating Tables II and III, and Figs. 2 and 3. The upper case letters in parentheses refer to Excel files in the SOM produced by the indicated step in the flow chart. However, (C) in the SOM represents an example file, i.e., SNP removal using the All SNPs(142) database. SOM, supporting online material; WXS, whole exome sequence; SNV, single nucleotide variant; SNP, single nucleotide polymorphism; USF, University of South Florida; CPCR, cytoskeleton-related protein coding regions; PROVEAN, Protein Variation Effect Analyzer.

Table I. HUGO symbols for the cytoskeletal protein-related coding regions set.

Table I.

HUGO symbols for the cytoskeletal protein-related coding regions set.

HUGO symbol	Gene name
ANK2	Ankyrin 2, neuronal
APC	Adenomatous polyposis coli
COL11A1	Collagen, type XI, α1
DNAH10	Dynein, axonemal, heavy chain 10
DNAH11	Dynein, axonemal, heavy chain 11
DNAH3	Dynein, axonemal, heavy chain 3
DNAH5	Dynein, axonemal, heavy chain 5
DNAH7	Dynein, axonemal, heavy chain 7
DNAH8	Dynein, axonemal, heavy chain 8
DSCAM	Down syndrome cell adhesion molecule
DST	Dystonin
FAT3	FAT atypical cadherin 3
FAT4	FAT atypical cadherin 4
FBN2	Fibrillin 2
FGFR1	Fibroblast growth factor receptor 1
FLG	Filaggrin
MUC16	Mucin 16, cell surface associated
MUC17	Mucin 17, cell surface associated
MUC4	Mucin 4, cell surface associated
NEB	Nebulin
NEFH	Neurofilament, heavy polypeptide
NF1	Neurofibromin 1
PCDH15	Protocadherin-related 15
PCDHAC2	Protocadherin alpha subfamily C, 2
PCDHGC5	Protocadherin gamma subfamily C, 5
PCLO	Piccolo presynaptic cytomatrix protein
PKHD1	Polycystic kidney and hepatic disease 1 (autosomal recessive)
PLEC	Plectin
RELN	Reelin
SPTA1	Spectrin, alpha, erythrocytic 1
SPTAN1	Spectrin, alpha, non-erythrocytic 1
SSPO	SCO-spondin
SYNE1	Spectrin repeat containing, nuclear envelope 1
SYNE2	Spectrin repeat containing, nuclear envelope 2
TTN	Titin
XIRP2	Xin actin binding repeat containing 2

[i] HUGO, Human Genome Organisation.

Use of SNP databases and the PROVEAN web tool

Two SNP databases [All SNPs(142) and 1000 Genomes; genome.ucsc.edu] were used to filter the previously identified variants in the raw sequence files representing the matched BRCA and normal blood barcodes, as indicated in Fig. 1. An example of this process is provided in two SOM files, ‘Stoll 2016 SOM Figure 1C (All SNPs(142) Example)’ and ‘Stoll 2016 SOM Figure 1C (All SNPs(142) Example, Excel)’. Variants not identified by either SNP database were classified by frequency. The classification of variants can be found in SOM Excel file, ‘Stoll 2016 SOM Figure 1E-BreastBlood Variants’ in sheet ‘Variant Color-Coding.’ Unique matched variants were defined as SNVs that were present in the BRCA and normal blood sample matched barcodes. A histogram of the number of cancer-/blood-matched CPCR SNVs limited to a single individual is presented in Fig. 2. These SNVs, as well as the non-unique SNVs identified among the investigated set of 31 individuals in this report (but not present in the SNP databases) served as input for the Protein Variation Effect Analyzer (PROVEAN) human genome variants program (http://provean.jcvi.org/genome_submit_2.php). This tool was used to determine if the matched mutations caused damaging amino acid changes. A summary of the PROVEAN results is presented in Table II. In addition, the SIFT analysis (http://provean.jcvi.org/genome_submit_2.php) of amino acid substitutions is available at the indicated web site and the results for the current project for the SIFT and PROVEAN analyses are present in the SOM files labeled ‘Stoll 2016 SOM Figure 1E-Provean Results, aa variants unique to a barcode’ and ‘Stoll 2016 SOM Figure 1F-Provean Results, aa variants unique to a barcode’.

Figure 2. Bar chart representing the number of cancer-/blood-matched CPCR SNVs limited to the indicated, individual barcodes (i.e., that are limited to a single individual). The horizontal axis represents The Cancer Genome Atlas (TCGA) barcodes while the vertical axis represents the number of cancer-/blood-matched CPCR SNVs present represented by the indicated barcode. CPCR, cytoskeletal protein-related coding regions; SNV, single nucleotide variant.

Table II. Deleterious amino acids identified in the cancer-/blood-matched cytoskeletal protein-related coding region single nucleotide variants and limited to one barcode (individual).

Table II.

Deleterious amino acids identified in the cancer-/blood-matched cytoskeletal protein-related coding region single nucleotide variants and limited to one barcode (individual).

Residue (wild-type)	Residue (mutant)	PROVEAN prediction	Gene name	Amino acid length
Barcode: AAAU
A	V	Deleterious	TTN	27118a
Barcode: AALI
G	C	Deleterious	FLG	4061
R	C	Deleterious	DST	5171
Barcode: A899
P	L	Deleterious	DNAH7	4024
Barcode: AALJ
H	Y	Deleterious	TTN	27118a
Barcode: A0SE
G	S	Deleterious	MUC4	1176
Barcode: A0CL
L	S	Deleterious	FAT3	892
Barcode: A0CP
P	T	Deleterious	TTN	33423a
Barcode: A0CS
E	K	Deleterious	TTN	27118a
Barcode: A0CT
K	E	Deleterious	DNAH10	4471
Barcode: A0CV
C	Y	Deleterious	RELN	3458
Barcode: A0CW
L	R	Deleterious	MUC16	14507
D	Y	Deleterious	ANK2	3924
Barcode: A0ER
R	C	Deleterious	SYNE2	6818
Barcode: A0EM
V	A	Deleterious	TTN	27118
Barcode: AALK
S	L	Deleterious	NEB	6669

a Differing lengths represent different protein IDs. PROVEAN, Protein Variation Effect Analyzer.

Statistical analysis

Statistical significance was determined using Student's t-test with Microsoft Excel (version 2010). P<0.05 was considered to indicate a statistically significant difference.

Results

SNVs unique to individuals of the BRCA study set and representing deleterious amino acids

To assess SNVs potentially present in CPCRs (Table I) of BRCA and normal blood samples of a single individual, a series of processing steps, indicated in Fig. 1 were performed. Any SNVs that were present in the All SNPs(142) and 1000 Genome SNP databases (Fig. 1) were eliminated from further consideration. In addition, any SNVs that were not verified by 20 reads or more were eliminated from consideration. Results indicated that 29 barcodes had CPCR SNVs, i.e., nucleotides not matching the reference human genome (Fig. 2). Of these, 14 barcodes represented deleterious amino acid replacements, according to PROVEAN analysis, in the following coding regions: TTN, MUC4, FLG, DST, DNAH7, FAT3, DNAH10, RELN, MUC16, ANK2, SYNE2 and NEB (Table II). The SIFT analysis indicated 21 damaging mutations (SOM Excel file labeled, ‘Stoll 2016 SOM Figure 1F-Provean Results, aa variants unique to a barcode’).

SNVs shared among members of the BRCA study set, but not present in common SNP databases

The results of the current SNV analyses indicated that a number cancer-/blood-matched CPCR SNVs were shared between barcodes in this analysis group of 31 barcodes, despite using the above two SNP databases to clear the known SNPs. Fig. 3 demonstrates the chromosome and nucleotide positions of the cancer-/blood-matched SNVs identified in more than one barcode, as well as the number of barcodes with a given variant. These SNVs were also analyzed using the PROVEAN tool (Table III; Stoll 2016 SOM Figure 1G-Provean Results, aa variants present in more than one barcode). A summary of the specific, non-unique CPCR SNVs is provided in Table III. Thus, 112 SNVs unique to one individual (Fig. 2) indicated 16 deleterious AA substitutions (Table II), whereas 29 SNVs shared among the barcodes represented zero deleterious AA substitutions, using the PROVEAN analysis (P<0.00002; Student's t-test). The SIFT analysis detected 21 deleterious AA among the unique, matched SNVs, and there were three deleterious AA substitutions among the shared, matched SNVs. The SIFT analysis maintains the trend indicated by the PROVEAN analysis, but was not statistically significant.

Figure 3. Bar chart representing the CPCR SNVs identified in multiple individuals that are not detected by All SNPs(142) or 1000 Genomes databases. The lower number on the horizontal axis indicates the chromosome number, while the upper number indicates the nucleotide position in the hg19 reference genome. The vertical axis represents the number of individuals that possess the indicated SNV. CPCR, cytoskeletal-related protein coding regions; SNV, single nucleotide variant; SNP, single nucleotide polymorphism.

Table III. Amino acid changes identified in the cancer-/blood-matched cytoskeletal protein-related coding region single nucleotide variants of multiple individuals.

Table III.

Amino acid changes identified in the cancer-/blood-matched cytoskeletal protein-related coding region single nucleotide variants of multiple individuals.

Residue (wild type)	Residue (mutant)	PROVEAN prediction	Gene name	Amino acid length
3,195505859; Barcodes: A0SE, A0CR
T	A	Neutral	MUC4	4442
3, 195507398; Barcodes: A89A, AALJ, A0SB, A0SE, A0CL, A0CR, A0CO, A0EW, A0CS
D	N	Neutral	MUC4	4442
3, 195507399; Barcodes: A0SB, A0SM, AALK, A0CV, A0CR, A0CO, A0EW
T	T	Neutral	MUC4	4442
3,195507406; Barcodes: A0CL, AALJ
P	L	Neutral	MUC4	4442
3, 195507422; Barcodes: A89A, A0SE, A0CL, A0CR, A0CO
D	H	Neutral	MUC4	4442
3, 195507443; Barcodes: AAAU, A89A, AALJ, A0SB, A0CL
T	P	Neutral	MUC4	4442
3,195507445; Barcodes: A0SB, A0CL, A0CS
D	V	Neutral	MUC4	4442
3,195508667; Barcodes: A89A, AALK, A0CR
T	A	Neutral	MUC4	4442
3,195508670; Barcodes: A89A, AALK, A0CR
D	H	Neutral	MUC4	4442
3,195510238; Barcodes: A0SB, AALK, A0CL
L	P	Neutral	MUC4	4442
3, 195510310; Barcodes: A89A, A0SB, A0CV, A0CO, A0CR, A0EW
Y	S	Neutral	MUC4	4442
3, 195510341; Barcodes: A89A, A0SB, A0SE, A0CW, A0CV, AALK, A0CO, A0CR, A0EW, A0CS
S	P	Neutral	MUC4	4442
3,195510610; Barcodes: A0SE, A0CR
L	P	Neutral	MUC4	4442
3, 195510611; Barcodes: A0SB, A0SE
L	I	Neutral	MUC4	4442
3, 195510613; Barcodes: A0SB, A0SE, A0CO
S	N	Neutral	MUC4	4442
3,195510622; Barcodes: A0CW, A0EW
P	H	Neutral	MUC4	4442
1, 103468336; Barcodes: A899, A89A, A6ZH, AALJ, A0CU, A0CT, A0EM, A0CM, A04Y, A04X, A0D0, A0EW
G	G	Neutral	COL11A1	1767

[i] PROVEAN, Protein Variation Effect Analyzer.

Discussion

The above data and analyses indicate that systemic mutations in CPCRs are relatively common, consistent with the large size of these coding regions. In general, systemic, de novo mutations are considered sporadic and unpredictable, as in the case of Rett's syndrome, which can be caused by a point mutation in methyl CpG binding protein-2 (6). However, the large size of the CPCRs renders it more likely that de novo, or relatively recently generated SNVs, will be present in the CPCRs. Furthermore, the above analyses indicate that SNVs that lead to deleterious amino acid substitutions are more common when not shared.

The relatively common occurrences of these de novo, systemic deleterious amino acid substitutions in CPCRs raises questions regarding the value of evaluating these coding regions. For example, small coding regions are unlikely to have de novo mutations and, thus, there may not be a practical justification for routine analysis. However, if de novo CPCR SNVs are identified to be essentially inevitable, investigating them would be valuable, particularly with regard to potentially identifying deleterious amino acid alternations. Such routine analysis may have value in better understanding a range of medical conditions, for example heart disease, due to the major role of various CPCRs in the formation of the sarcomere cytoskeleton (7).

In addition, the results cause us to query other coding region groups, where individual group members may be too small to regularly reveal systemic mutations. The entire group collection, however, may be large enough such that de novo systemic mutations would alter one member of the group often enough to justify screening. For example, a comprehensive collection of tumor suppressor genes would likely have a member that represents a systemic mutation, due to random chance, fairly often. While such a collection is not likely to be as large as a CPCR collection, it is likely a large enough portion of genome, with mutations affecting a large enough number of individuals, to justify screening. Furthermore, due to the degeneracy of numerous tumor suppressor signaling pathways (8), a mutation in any one tumor suppressor, among a set, may represent pre-disposition to cancer.

In conclusion, SNPs not present in the SNP databases used here, but that were present in more than one of the individuals investigated above, and potentially present in other SNP databases, were detected. However, as all of the individuals analyzed in the current study were BRCA patients, the outcome of multiple shared SNVs results in the hypothesis that certain de novo, systemic CPCR mutations may facilitate BRCA development. Future studies are required to address the occurrence of potential, cancer specific SNVs in the CPCR set by repeating the above analyses for multiple types of cancer.

Acknowledgements

Authors would like to acknowledge support from the Anna Valentine program and the University of South Florida research computing facility, in particular, Dr Tony Green.

Glossary

Abbreviations

Abbreviations:

References