Panel-based testing for inherited colorectal cancer: a descriptive study of clinical testing performed by a US laboratory

Next-generation sequencing enables testing for multiple genes simultaneously (‘panel-based testing’) as opposed to sequential testing for one inherited condition at a time (‘syndrome-based testing’). This study presents results from patients who underwent hereditary colorectal cancer (CRC) panel-based testing (‘ColoNext™’). De-identified data from a clinical testing laboratory were used to calculate (1) frequencies for patient demographic, clinical, and family history variables and (2) rates of pathogenic mutations and variants of uncertain significance (VUS). The proportion of individuals with a pathogenic mutation who met national syndrome-based testing criteria was also determined. Of 586 patients, a pathogenic mutation was identified in 10.4%, while 20.1% had at least one VUS. After removing eight patients with CHEK2 mutations and 11 MUTYH heterozygotes, the percentage of patients with ‘actionable’ mutations that would clearly alter cancer screening recommendations per national guidelines decreased to 7.2%. Of 42 patients with an ‘actionable’ result, 30 (71%) clearly met established syndrome-based testing guidelines. This descriptive study is among the first to report on a large clinical series of patients undergoing panel-based testing for inherited CRC. Results are discussed in the context of benefits and concerns that have been raised about panel-based testing implementation. Conflict of interest Cristi Radford and Jill Dolinsky are full-time employees for the commercial laboratory Ambry Genetics, which performs ColoNext™ testing. Elizabeth Chao is a paid consultant for Ambry. Deborah Cragun, Meghan Caldwell, and Tuya Pal report no potential conflicts of interest. Specifically, they are not employed by Ambry, and they did not receive any financial or other incentives from Ambry.

DNA sequencing technologies (called next-generation sequencing; NGS) now make it possible to test for multiple genes simultaneously (panel-based testing), at a cost comparable to testing for two genes using older methods (syndrome-based testing) (5). As a result, it is possible that the conventional syndrome-based approach to performing inherited cancer predisposition testing, which includes generating a differential diagnosis and sequentially testing for single genetic conditions, may shift to panel-based tests.
In March 2012, Ambry Genetics Corporation (Ambry; Aliso Viejo, CA) was the first clinical laboratory to offer hereditary cancer panel-based testing in the United States; subsequently, other laboratories have started offering cancer panels. The 14 genes included in Ambry's colon panel-based testing (ColoNext ™ ) are listed in Table 1 along with their associated cancer risks. Although hereditary cancer panels vary, they typically include both highly penetrant as well as moderately penetrant genes (6). For highly penetrant genes, clinical guidelines exist for the prevention or early detection of cancers (7). In other words, these are 'actionable genes' with known clinical utility (8). In contrast, the utility of moderately penetrant genes is less certain.
As panel-based testing is implemented into clinical practice, cost-saving opportunities need to be considered. In particular, the most common cause of hereditary CRC is Lynch syndrome, which is caused by mutations in five genes (i.e. MLH1 , MSH2 , MSH6 , PMS2 , EPCAM ). When this study was initiated, syndromebased testing for the Lynch syndrome genes through Ambry cost $100 more than ColoNext ™ . However, pricing has fluctuated and ColoNext ™ testing has become more expensive than testing for the five Lynch syndrome genes. Cost analyses are further complicated because tumor screening for Lynch syndrome using immunohistochemical (IHC) testing may narrow down the number of genes that require testing (9).
Despite the potential for panel-based testing to identify more mutations, this testing is also expected to increase the complexity of results interpretation because of factors such as questionable or uncertain clinical utility of testing for moderate penetrance genes and the higher rate of inconclusive results because of an increase in the number of variants of uncertain significance (VUS) (6,10,11). Furthermore, given that widespread panel-based testing for hereditary CRC only began in 2012, little is known about patients who are tested or rates of identified mutations and variants. To address these questions, we examined a comprehensive data repository of panel-based testing for inherited susceptibility to CRC that is maintained through Ambry. Unlike prior validation studies of cancer panel-based tests (5,12), the purpose of this study was to describe the clinical use and results of ColoNext ™ testing performed on a clinical basis for diagnostic purposes in patients without previously identified mutations. The aims of this study were to estimate mutation and VUS rates identified through panel-based testing for hereditary CRC in real-world settings and to determine whether patients with a mutation met national genetic testing criteria for the respective cancer syndromes identified.

Data source
A database maintained by Ambry includes demographic information as well as personal and family cancer history collected from ordering clinicians via test requisition forms (TRF). Data extracted by Ambry for use in this study included de-identified information on individuals for whom ColoNext ™ testing was completed between March 2012 (when ColoNext ™ became available) and March 2013. Following receipt of an exempt certification by the University of South Florida's institutional review board, the first and senior authors (D. C. and T. P.) conducted secondary data analysis on the de-identified dataset.
ColoNext ™ testing and results reporting All genetic testing was performed by Ambry using the following protocol. Genomic deoxyribonucleic acid (gDNA) was isolated from patients' whole blood specimens, or from saliva specimens collected using an Oragene kit. Sequence enrichment of each coding exon within the 14 genes was carried out by incorporating the gDNA into microdroplets along with primer pairs designed to the specified target. The enriched libraries were then applied to the solid surface flow cell for clonal amplification and sequencing using paired-end, 100-cycle chemistry on the Illumina Hiseq 2000 (Illumina, San Diego, CA). For 13 of the 14 cancer susceptibility genes, NGS/Sanger sequencing was performed for all coding domains plus at least five bases into the 5 and 3 ends of all introns and untranslated regions. Sequence analysis was not performed for EPCAM , as currently the only mutations in EPCAM associated with Lynch syndrome are gross deletions encompassing its 3 terminus (13). Additional Sanger sequencing was performed for any regions with insufficient depth of coverage by NGS, with an initial read depth threshold of at least 10 times and a quality score of 20 or better. This threshold was later increased to a read depth of 50 times with a quality score of 20 or better. Variant calls other than known, non-pathogenic alterations were verified by Sanger sequencing in sense and antisense directions prior to reporting. A targeted chromosomal microarray (CMA) designed with increased probe density in regions of interest was used for the detection of gross deletions and duplications for each sample (Aglient, Santa Clara, CA). Owing to potential pseudogene interference (14), PMS2 sequence analysis was performed via long range PCR followed by Sanger sequencing and PMS2 deletion/duplication analysis was performed via MLPA. If a gross deletion was detected in exons 12, 13, 14, or 15 of PMS2 , double stranded sequencing of the appropriate exon(s) of the pseudogene was performed to determine if the deletion was located Personal and family history information was manually reviewed and coded according to the presence or absence of colon cancer, endometrial cancer, other cancers, and colon polyps. Ordering provider information was coded to reflect involvement of a genetic professional (i.e. board certified/eligible medical geneticist or genetic counselor) for cases meeting one of the following criteria: (1) a master's trained genetic counselor or medical geneticist was listed on the TRF; or (2) in cases where no genetic professional was listed, a search of Ambry's internal database of providers and publically available websites revealed that the ordering provider works closely with a genetic counselor. In order to characterize the population of patients who underwent testing, frequencies for available demographic, clinical, family history, and healthcare provider variables were calculated. Frequencies were also calculated after subgrouping according to a positive test result, and further sub-dividing according to whether the positive result was actionable.
Personal and family history information for all patients with an actionable result was reviewed by the first author (D. C.) and independently verified by at least one of the other study co-authors (T. P. or J. D.) to determine whether individuals met the respective NCCN syndrome-based testing guidelines presented in Table 1. Finally, available clinical and family history information was summarized for all patients with a CHEK2 mutation.

Demographic and clinical characteristics
Our series consisted of 586 individuals who underwent ColoNext ™ testing between March 2012 and March 2013. Testing was ordered at 216 unique institutions across the US. The majority of individuals tested were female (60%) and White (72%). Just over half (n = 311; 53%) had a personal history of CRC (with or without a history of other cancers or polyps), 105 (18%) had a history of cancer other than CRC, and 123 (21%) had a history of polyps with no personal cancer history. The majority (n = 316; 54%) had a positive family history of CRC/or other cancers. Most tests (89%) were ordered by genetic professionals or physicians who work closely with a genetic counselor. Provider settings were diverse, including private offices, community hospitals, and academic institutions. When sub-dividing by test result (i.e. positive and actionable), demographic and clinical characteristics remained similar across the groups ( Table 2).

Pathogenic and actionable mutations
Results of ColoNext ™ testing are summarized in Fig. 1. Of 586 patients, 61 (10.4%) were positive for a mutation in at least one of the genes analyzed. Only one person had two pathogenic mutations (MLH1 and CHEK2 ). After removing the 8 other patients with CHEK2 mutations and all 11 patients in whom only 1 MUTYH mutation was identified, the number of patients with actionable mutations decreased to 42 (7.2%), with over half of these occurring in Lynch syndrome genes.

Variants of unknown significance (VUS)
A total of 118 individuals (20.1%) had at least one VUS identified. These included: 14 with a pathogenic mutation and one or more VUS; 99 with one VUS; and 15 with more than one VUS (Fig. 1). Among all 159 VUS results, 77 (48%) occurred in one of the genes associated with Lynch syndrome.

Syndrome-based testing/screening guidelines
As shown in Fig. 1, of the 42 patients with an actionable mutation, 30 (71%) clearly met NCCN syndromebased testing, screening, or diagnostic criteria listed in Table 1. Available information for those 12 not clearly meeting criteria is included in Table 3. Table S1, Supporting Information summarizes information on the eight patients identified with CHEK2 mutations. All of these individuals had a personal history of adenomatous polyps or CRC. None of these individuals had a personal history of breast cancer, but six had at least one family member with breast cancer.

Discussion
To our knowledge, this study is among the first to report on a large clinical series of patients tested for inherited CRC through a NGS panel-based test. did not include juvenile polyps and in the other case no history of polyps was reported. Additionally, there were four cases in whom mutations in polyposis genes were detected (i.e. two MUTYH and two APC ); yet these individuals had insufficient numbers of polyps to meet NCCN testing criteria for polyposis syndromes. These findings highlight the potential for panel-based testing to identify mutations that might otherwise not have been identified because of limited medical or family history or an atypical presentation. Of note, there were several cases where criteria for a specific genetic syndrome were clearly met in which there was an overlapping phenotype with other genetic conditions [e.g. CRC is associated with Lynch syndrome, but a similar phenotype may be observed with attenuated familial adenomatous polyposis (FAP)] (17). These cases represent situations where panelbased testing (compared to syndrome-based testing) may be particularly useful. Indeed, over half of the actionable mutations identified in this study occurred in genes associated with Lynch syndrome and several individuals with positive results had features of more than one condition.
As panel-based testing only became clinically available in March 2012, few reports exist to which our findings may be compared. One recent institution-based series of 50 patients with clinical panel-based testing for inherited cancer through Ambry included only five patients who received ColoNext ™ testing (18). Of these, one had a CHEK2 mutation and another was a MUTYH heterozygote.
Despite limited published data on clinical outcomes of panel-based testing, its potential benefits in the context of hereditary CRC have been reported (5,6). For example, panel-based testing can be less time consuming than syndrome-based testing (6). However, the cost-efficiency of panel vs syndrome-based testing will ultimately be determined by the relative costs of each, and costs continue to evolve within the current testing landscape. On the basis of our findings, although there are scenarios where panel-based testing may have been more cost-efficient, reality remains that syndromebased testing would have been sufficient to identify the majority of patients with deleterious mutations. Consequently, the optimal and most cost-effective use of panel-based testing as a first-tier test vs a secondtier test (i.e. after syndrome-based testing is negative), remains to be determined. Although data suggest that several individuals in this study previously underwent syndrome-based testing, it is unknown whether these represent second-tier testing because initial testing could have been performed prior to the availability of panel-based testing. Furthermore, as information on tumor testing for Lynch Syndrome (either through IHC and/or MSI) was not included in the majority of TRFs, it was not possible for us to assess the efficacy of tumor testing to inform germline testing in our sample. Specifically, tumor testing (either through IHC and/or Table 3. General descriptions of individuals with a positive 'actionable' mutation a who did not clearly meet the testing criteria listed in Table 1 Gender Gene and mutation MSI) was mentioned in 21 cases. Of these, four had either abnormal IHC and/or MSI-H tumors, of which one had a mutation in MLH1 and the other three had prior testing for one or two Lynch syndrome genes. The remaining 17 cases for which this data was available through the TRF had normal IHC and/or MSI tumor testing.
Despite potential benefits, several anticipated challenges have been cited when conducting panel-based testing for inherited cancer predisposition (6,18). For example, testing for mutations in moderate penetrance genes could potentially lead to increased patient anxiety and/or excessive and unnecessary screening or preventive surgeries. Furthermore, there may be increased time for counseling when performing panelbased testing, in part because of the higher VUS rate and relaying information to patients about moderate penetrance genes. Nevertheless, as more individuals with mutations in moderate risk genes are identified, because of their inclusion on panel-based tests, additional information about their implications for cancer risk and medical management is expected to emerge.
The likelihood of detecting mutations in moderate penetrance genes is dependent on which of these genes is included on the specific panel-based test ordered. In our report focused on a colon-specific panel, CHEK2 is the only moderate penetrance gene included within this panel. Different mutations in CHEK2 may be associated with substantially different cancer risks (19). Clinical testing for CHEK2 has been available for years, yet medical management of patients with CHEK2 mutations is not well-defined; and cancer screening recommendations are primarily based on patients' personal and family medical histories. Individuals with CHEK2 mutations in this study may have a higher incidence of CRC because of selection bias. However, prior to the introduction of panel-based testing, clinical testing for CHEK2 mutations was rare, and our results may reflect that these mutations are more common than previously believed. While most cancers in the eight CHEK2 families were later onset, seven of them had at least one individual with a cancer diagnosed before the age of 50. In these cases it may be appropriate to consider lowering the age they begin cancer screening for the affected organ (19).
Additional challenges can occur related to MUTYH associated polyposis (MAP) which confers up to 63% CRC risk by age 60 (20). Unlike other hereditary CRC syndromes that are included on ColoNext ™ testing, MAP is an autosomal recessive condition; therefore, parents and children of individuals with MAP are rarely affected. Nevertheless, the question of whether monoallelic MUTYH mutation carriers have a moderately increased CRC risk remains unclear (20)(21)(22).
Another factor to consider with expansion of panelbased testing is an expected increase in the rate of VUS results. This study found a relatively high VUS rate, occurring in 20% of the individuals tested. Interestingly, nearly half of all VUS results in the current study occurred in four Lynch syndrome genes which comprised less than a third of the 14 ColoNext ™ genes, which translates to a VUS rate of ∼10% consistent with reports from other laboratories that perform LS genetic testing (23,24). Although VUS results remain an important concern, it is anticipated that the ability to classify VUS results will improve with more widespread testing and data sharing among researchers and laboratories (including Ambry).
Study strengths include the large sample size from the first commercial laboratory to offer clinical panel-based testing for inherited cancer predisposition. Furthermore, generalizability of findings is enhanced given that our sample included a diverse group of patients from over 200 US institutions. Despite these strengths, our sample encompassed early adopters of a single cancer panel primarily ordered by genetics professionals. Therefore, mutation and VUS rates will vary across panels and may change over time as panel-based tests are more widely diffused. Furthermore, our reliance on information in the TRF without medical record and family history verification as well as lack of information about tumor testing preceding germline testing in most cases (to identify possible Lynch Syndrome) are inherent limitations; however this approach also enabled the large sample size for the study. Moreover, information contained in the TRF was sufficient to suggest that phenotypes associated with mutations in certain genes and criteria for testing may need to be expanded as more panel-based testing identifies other atypical cases.
As adoption of panel-based testing for hereditary cancer syndromes continues to spread it is important to ensure that patients receive accurate information about their test results, particularly for the substantial number of patients who will receive inconclusive results as well as those found to have mutations in rarer inherited cancer genes and moderate penetrance genes. This study suggests that genetics professionals comprise most of the early adopters of panel-based testing. However, it is anticipated that over time, the use of NGS panels will diffuse to providers with less specialized training in genetics. Educational efforts, preferably spearheaded by academic institutions or professional organizations rather than by commercial laboratories, will therefore be needed as many providers in the US have been shown to lack knowledge of fundamental genetic concepts and VUS results (25)(26)(27). Furthermore, although panelbased testing may reduce the need to perform a differential diagnosis up front, clinicians who order testing must still be knowledgeable about clinical diagnostic criteria for hereditary cancer syndromes and management recommendations in order to provide the most appropriate application of test results to patient care. Finally, it remains uncertain whether identification of moderate penetrance genes truly helps guide cancer screening decisions over and above what would be recommended based on comprehensive collection of family history without conducting testing.
Ultimately, to help prevent negative outcomes, it is imperative that additional guidelines regarding the use of panel-based testing be developed. The current not mention panel-based testing. Additional research is needed to update these practice guidelines so they may better address the unique practice-based issues and advantages of panel-based testing. Uncertainties associated with moderate penetrant genes, such as CHEK2 , highlight the importance of research studies and academic registries. In addition, research is needed to identify the optimal counseling approach for panelbased testing.
Overall, this study provides a broader picture of panel-based testing for hereditary CRC than previously available and suggests that testing may: (1) cast a wider net, and in some cases identify mutations in genes that might not otherwise be tested because of an atypical phenotype and (2) be an efficient approach when patients present with features of more than one hereditary CRC syndrome. Despite highlighting these potential benefits of panel-based testing, this study raises additional questions and concerns that will need to be addressed through clinical research and education.

Supporting Information
The following Supporting information is available for this article: Table S1. Descriptions of individuals with a pathogenic mutation in CHEK2 only.
Additional Supporting information may be found in the online version of this article.