Clinical Correlates of Autosomal Chromosomal Abnormalities in an Electronic Medical Record–Linked Genome-Wide Association Study

Although mosaic autosomal chromosomal abnormalities are being increasingly detected as part of high-density genotyping studies, the clinical correlates are unclear. From an electronic medical record (EMR)–based genome-wide association study (GWAS) of peripheral arterial disease, log-R-ratio and B-allele-frequency data were used to identify mosaic autosomal chromosomal abnormalities including copy number variation and loss of heterozygosity. The EMRs of patients with chromosomal abnormalities and those without chromosomal abnormalities were reviewed to compare clinical characteristics. Among 3336 study participants, 0.75% (n = 25, mean age = 74.8 ± 10.7 years, 64% men) had abnormal intensity plots indicative of autosomal chromosomal abnormalities. A hematologic malignancy was present in 8 patients (32%), of whom 4 also had a solid organ malignancy while 2 patients had a solid organ malignancy only. In 50 age- and sex-matched participants without chromosomal abnormalities, there was a lower rate of hematologic malignancies (2% vs 32%, P < .001) but not solid organ malignancies (20% vs 24%, P = .69). We also report the clinical characteristics of each patient with the observed chromosomal abnormalities. Interestingly, among 5 patients with 20q deletions, 4 had a myeloproliferative disorder while all 3 men in this group had prostate cancer. In summary, in a GWAS of 3336 adults, 0.75% had autosomal chromosomal abnormalities and nearly a third of them had hematologic malignancies. A potential novel association between 20q deletions, myeloproliferative disorders, and prostate cancer was also noted.


Introduction
Genome-wide association studies (GWAS) are identifying novel genomic loci associated with various diseases and quantitative traits. 1,2 Chromosomal abnormalities such as mosaic deletions, amplifications, and unipaternal disomies are incidentally found on signal intensity analyses in such studies. The clinical correlates of these abnormalities remain poorly defined. 3 In 2007, the National Human Genome Research Institute funded the electronic MEdical Record and GEnomics (eMERGE) consortium to study and evaluate the utility of high throughput electronic medical record (EMR)-based phenotyping methods to facilitate genomic studies. 4,5 This approach was successfully applied to GWAS of several quantitative traits including red blood cell indices, erythrocyte sedimentation rate, white blood cells, and PR interval of the electrocardiogram. 4,[6][7][8][9] The availability of high-density genotyping data linked to the EMR in the eMERGE consortium offers an opportunity to study the clinical correlates of incidentally found chromosomal abnormalities. The GWAS at Mayo Clinic was conducted to identify loci associated with peripheral arterial disease (PAD). An important area of investigation involves return of incidentally found genetic abnormalities. As a step in this direction, we conducted a detailed EMR review of patients with incidentally found autosomal chromosomal abnormalities and controls without such abnormalities, to ascertain the clinical correlates of these abnormalities.

Materials and Methods
The study protocol was approved by the Mayo Clinic Institutional Review Board and included 3336 participants: 1687 PAD cases (mean age = 65.8 ± 10.7 years, 64.3% men) and 1649 control subjects (mean age = 60.6 ± 7.3 years, 59.8% men). Written informed consent was obtained. Details of patient recruitment and characteristics have been previously described. 4,6 Genotyping was performed at the Center of Genotyping and Analysis at the Broad Institute, Cambridge, Massachusetts, using the Illumina Human 660W-Quad V1 genotyping platform that consists of 561 490 single nucleotide polymorphisms (SNPs) and 95 876 intensity-only probes.
A genomic data analysis pipe-line that combined 3 GenomeStudio plug-ins (cnvPartition 1.2.1, LOH detector, and ChromoZone; Illumina, San Diego, CA) waas run using log-R ratio (LRR) and B-allele frequency (BAF) data to identify copy number variation (CNV) and loss of heterozygosity (LOH) regions (both copy neutral LOH and heterozygous deletion LOH). CnvPartition 1.2.1 was used to estimate copy number and for annotation of chromosomal regions with CNV. LOH detector was used to detect extended tracts of homozygosity and ChromoZone was used to auto-bookmark for single-sample analysis. The computational genomics pipeline used to define a chromosome abnormality is summarized in Figure 1, and the computational script can be found in the supplementary material (available online at http://HIC.sagepub.com/supplemental). The unions of LOH and CNV from these 3 algorithms were combined using a Perl script developed in-house. A chromosomal abnormality was considered to be present when the combined abnormalities exceeded 20% of the chromosome's total length. The signal intensity data from each abnormal chromosome were visually examined to validate the call. We used previously established definitions 10 in classifying the observed chromosomal abnormalities. Deviations from the expected normal bipaternal disomic state were assessed using LRR and BAF data. In a normal study participant, BAF at any locus is Three GenomeStudio plug-ins (cnvPartition 1.2.1, LOH detector, and ChromoZone; Illumina, San Diego, CA) were combined and run using log-R ratio (LRR) and B-allele frequency (BAF) data to identify copy number variation (CNV) and loss of heterozygosity (LOH) regions (both copy neutral LOH and heterozygous deletion LOH). The unions of LOH and CNV from these 3 algorithms were combined using a Perl script developed in-house. A chromosomal abnormality was considered to be present when the combined abnormalities exceeded 20% of the chromosome's total length. The signal intensity data from each abnormal chromosome were visually examined to validate the call. expected to be either 0 (AA), 0.5 (AB), or 1 (BB) corresponding to an LRR of 0. Negative deviations of LRR corresponded to deletions whereas positive deviations were compatible with amplifications. Samples with BAF asymmetry and LRR of 0 were classified as unipaternal disomies (UPD).
The EMR of 25 patients with chromosomal abnormalities and 50 randomly selected age-, sex-, and PAD casematched participants without chromosomal abnormalities (in a 1:2 ratio) were reviewed in detail by one of the authors (HJ) to ascertain associated disease states and other clinical characteristics. The review included all physician notes, radiology studies, laboratory results, and pathology reports.
Five patients had chromosome 20q deletions (Ch20q del), and we identified the common deleted region in these patients. Using the break points for each of the 5 patients, the deleted segments of Ch20q were mapped to the human reference sequence (NCBI36/hg18) in UCSC Genome browser using custom tracks. The genes encoded in the common deleted region were identified from NCBI36/Ensembl54 database using the Bioconductor package bioMart. 11

Results
Of the 3336 genotyped study participants, 25 patients (PAD cases: n = 20, controls: n = 5, mean age = 74.8 ± 10.7 years, 64% male, 0.75% of total study participants) were found to have abnormal intensity plots consistent with the following chromosomal abnormalities: mosaic deletions (DEL = 7), mosaic and typical UPD (UPD = 18), and one chromosome with both amplification/UPD. One patient had 2 abnormalities: UPD and DEL. Table 1 lists the type of the chromosomal abnormality as well as the accompanying significant medical conditions ascertained by detailed review of the EMR. Supplementary material includes intensity plots of all patients in this report.
Of the 50 age-, sex-, and PAD case-matched participants without chromosomal abnormalities, a hematologic malignancy was present in only 1 patient, a significantly lower prevalence of hematologic malignancies than in patients with chromosomal abnormalities (2% vs 32%, respectively, P < .001). Although solid organ malignancies were slightly less common in patients without chromosomal abnormalities (n = 10, 20%) compared to patients with chromosomal abnormalities (n = 6, 24%), this difference was not statistically significant (P = .69). Solid organ malignancies in the 50 age-, sex-, and PAD case-matched participants without chromosomal abnormalities included prostate cancer (n = 7), colon cancer (n = 1), renal cell carcinoma (n = 1), and endometrial adenocarcinoma (n = 1).
Of 5 patients (3 men and 2 women) with overlapping mosaic deletions of Ch20q, 4 had a myeloproliferative disease (MPD) as mentioned above (polycythemia vera [n = 2], essential thrombocythemia [n = 1], and myeloproliferative/ myelodysplastic syndrome, unclassified [n = 1]). All men in this group had a history of prostate cancer in addition to MPD. One of the women in this group with MPD had history of prostate cancer in her father. The only patient without MPD had a persistently elevated erythrocyte sedimentation rate of unknown etiology and severe PAD. Of 18 patients with UPD, 14 had atherosclerotic vascular diseases (coronary artery disease, PAD, or stroke). Three of the patients with UPD had a hematologic malignancy whereas 4 had a solid organ malignancy. Two women with UPD of Ch22q had severe premature atherosclerosis necessitating revascularization for PAD at ages 47 and 52. One patient with deletion of Ch7 had chronic myelomonocytic leukemia and cytogenetic analysis (performed as a part of clinical evaluation) had confirmed Ch7 monosomy. Table 1 summarizes the major clinical diagnoses for all patients with chromosomal abnormalities in our GWAS.

Discussion
Given the proliferation of high-density genome-wide genotyping studies, an increasing number of mosaic autosomal chromosomal abnormalities are being detected. However, the clinical correlates of these abnormalities in study participants are not yet fully characterized. In the present study, we leveraged an EMR-linked GWAS of 3336 adults to identify the major clinical conditions associated with these abnormalities. We found that 0.75% of the study participants (n = 25) had autosomal chromosomal abnormalities. Linkage to the EMR enabled ascertainment of clinical features of these 25 patients beyond PAD case-control status. Compared to 50 age-, sex-, and case-matched participants without chromosomal abnormalities, hematologic malignancies were more frequent among patients with chromosomal abnormalities (32% vs 2%, P < .001). Solid organ malignancies were slightly more frequent in patients with chromosomal abnormalities but the difference was not statistically significant (24% vs 20%, P = .69). In 2 large studies 10,12 assessing chromosomal abnormalities in GWAS, such abnormalities were present in 0.80% to 0.89% of participants. The first of these 12 evaluated 31 717 cancer cases and 26 136 controls and found 517 patients (0.89%) had at least 1 chromosomal abnormality. Among patients with hematologic malignancies, nearly 20% had a chromosomal abnormality compared to 0.76% in cancer-free controls. These abnormalities were also associated with an increased rate of solid organ malignancies compared to cancer-free patients. Older age was associated with increased incidence of these abnormalities (0.23% for patients <50 years compared to 1.91% for patients 75-79 years old). 12 The second study 10 included more than 50 000 participants and demonstrated a 10-fold higher odds for a hematologic malignancy among patients with mosaic chromosomal abnormalities versus nonmosaic individuals. These abnormalities increased in prevalence with increasing age and were also associated with hematologic and solid organ malignancies. Our results are similar with regard to the higher prevalence of hematologic malignancies although the frequency of solid organ malignancies was not significantly different, likely due to the smaller sample size in our study.
A potential novel finding of our study is a possible association between Ch20q deletion, MPD, and prostate cancer. A myeloproliferative disorder was present in 4 (of 5) of these patients and prostate cancer in all 3 male patients. Interstitial deletions of Ch20q have been previously described in MPD and to a lesser extent in myelodysplastic syndromes. 13 Although Ch20q deletions are not pathognomonic for polycythemia vera (PV), up to 10% of PV patients have Ch20q deletions compared to 5% in patients with myelodysplastic syndrome. 14 Among patients with myelofibrosis (primary or secondary to PV and essential thrombocythemia), Ch20q deletion was the most frequent cytogenetic abnormality, occurring in up to 36% of affected patients. 15 Ch20q deletions have also been reported to be associated with acute lymphocytic and lymphoblastic leukemias. 16 The tumorigenic pathways underlying Ch20q deletions are still being investigated. Ch20q deletions may lead to loss of tumor suppressor genes and thereby increase the risk for developing hematologic malignancies. 17,18 The common deleted region in MPD patients was evaluated in several previous studies and a variety of break points in the region were noted with no homozygous deletion in Ch20q characterized till date. 10,19 Another group of researchers 20 evaluated the common retained regions of Ch20q in addition to the previously described common deleted region. They hypothesized that these common retained regions may contain oncogenes that may be overexpressed following Ch20q deletion and thereby contribute to the development of MPD. Other studies suggested that Ch20q deletion and subsequent genomic rearrangement may affect retained neighboring genes and result in either inhibition of tumor suppressor genes (such as DIDO1) or overexpression of other retained candidate oncogenes. [20][21][22] Several linkage studies demonstrated that 20q13 locus was associated with prostate cancer. [23][24][25][26] Ch20q gains have also been demonstrated to be associated with several malignancies including prostate cancer. [27][28][29][30] However, Ch20q deletions were less frequently associated with prostate cancer compared to Ch20q gains. 31,32 Considering that MPD and prostate cancer were associated with Ch20q deletions in our patients, altered gene expression of the retained regions of Ch20q is a more likely explanation for the observed phenotype. Ch20q deletion seems to result in either inhibition of tumor suppressor gene(s) or overexpression of oncogene(s) of the common retained regions of Ch20q. Identification of genes and pathways that might potentially lead to prostate cancer and MPD in the setting of Ch20q deletion requires further investigation.

Psychosocial and Ethical Implications
The incidental finding of autosomal chromosomal abnormalities in a GWAS has significant psychosocial and ethical implications. For example, one of the patients with Ch20q deletions did not have a diagnosis of malignancy in EMR but was noted to have persistently elevated erythrocyte sedimentation rate ranging between 35 and 50 mm/1 hour as well as a marginally elevated neutrophil count (5.6 × 10 9 /L to 8.3 × 10 9 /L), raising the concern that the patient could have an underlying malignancy or was at potential risk of developing malignancy in the future. At present, there are no prospective data that provide the relative risk for future hematologic or solid organ malignancies in otherwise asymptomatic patients with chromosomal abnormalities. There is also no consensus on how to appropriately manage such incidental findings nor any standardized protocols for synthesizing, analyzing, and disclosing this genetic information in the clinical setting. 33,34 Furthermore, whether the reportable information would ultimately be "actionable" or linked to downstream screening, diagnostic studies, or treatments remains uncertain. In the example described above, the incidental finding was not reported due to the paucity of correlative clinical data and the lack of evidence supporting downstream implications of Ch20q deletions as well as other genetic abnormalities. Additionally, informed consent (including for this study) often does not specify returning results to study participants. 35 Where disclosure is required, the use of laboratory tests that have been certified by Clinical Laboratory Improvement Act/Amendment is recommended to ensure federal regulations and proficiency standards are upheld.
We expect the number of incidentally found chromosomal abnormalities to increase as more GWAS are carried out. Prospective follow-up of patients with such chromosomal abnormalities will be required to ascertain outcomes including the risk of developing malignancy. Only then will we be able to understand how these abnormalities affect otherwise asymptomatic patients and whether further medical intervention is justified in these patients or not. Until a more robust understanding of the clinical relevance of genetic variants is obtained, the implications of these discoveries will remain unclear as would the investigator's obligation to disclose those findings to study subjects. Nonetheless, research study participants' demand for insight into these findings will likely increase as will the demand for disclosure. 36

Conclusion
In a GWAS of 3336 adults, 0.75% (n = 25) had autosomal chromosomal abnormalities, and of these 40% had a hematologic and/or a solid organ malignancy. Chromosomal abnormalities were significantly associated with the presence of hematologic malignancies. We also noted a possible association of Ch20q deletions with MPD and prostate cancer highlighting the potential of EMR-linked GWAS to uncover new genotype-phenotype correlations. Further research using larger cohorts will be required to confirm this finding. As the number of incidentally found chromosomal abnormalities in otherwise asymptomatic patients is expected to increase, there is a pressing need for prospective studies that evaluate the outcomes and downstream implications of such abnormalities.

Authors' Note
Supplemental data includes intensity plots of all detected autosomal chromosomal abnormalities and a table of genes within the common deleted region among the 5 patients with Ch20q deletions. The computational script can be found in the supplementary material in .PL format. Please view the README document for further instructions on the use of the Perl script used in this study.

Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.