It was just over a year ago, during the Association of Genetic Technologists' annual meeting, that Jun Gu presented a talk intended to be a subtle wake-up call to the cytogenetics community. Gu, an assistant professor at MD Anderson Cancer Center, urged his fellow cytogeneticists who were not using comparative genomic hybridization arrays to get with the program and avail themselves of the technology, which is quickly overtaking older techniques like karyotyping.
"I'm a cytogenetic-oriented person, so I see my task as trying to get other cytogeneticists aware of array CGH technology and use it more frequently," Gu says.
Consortia in the community have made similar proclamations. Not long after Gu's plea to his peers, the Association for Molecular Pathology informed the US Food and Drug Administration at a public meeting that array-based cytogenetic tests provide much higher resolution than traditional karyotyping. The International Standards for Cytogenomic Arrays Consortium has also suggested that chromosomal microarrays replace G-banded karyotyping and become the new clinical standard for diagnosing unexplained developmental disorders.
Gu and his colleagues published a paper in May in the American Journal of Medical Genetics demonstrating the technology's effectiveness at identifying complex rearrangements that traditional karyotyping failed to detect in a patient with several congenital anomalies. Using Agilent's 244K Whole Human Genome Oligo Microarray Kit, the team detected novel genomic rearrangements as mechanisms of constitutional defects on chromosome 16p13.3, involving a de novo deletion-triplication-duplication pattern that was confirmed by fluorescence in situ hybridization analyses.
"This is one application of array CGH in the [prenatal or postnatal] cytogenetics arena that will hopefully trigger more discussion about how this type of complex abnormality forms," Gu says. "At this time, more and more disease-specific array CGH chips are being developed, but they are still waiting to be validated."
Tackling tumors
As array CGH continues to make an impression in clinical diagnostics, the technology is also providing disease researchers with a way to rapidly identify potential target oncogenes and tumor suppressor genes for several common cancer types. The University of California, Los Angeles School of Medicine's Marlena Fejzo turned to array CGH to analyze the 20q13 stretch of the genome — one of the most frequent regions of copy-number abnormality in ovarian cancer. From that analysis, she found that ADRM1 was highly correlated with amplification among 20q13-amplified genes in ovarian cancer, and was the most significantly up-regulated gene in terms of metastasis, recurrence, and stage. Recently, Fejzo used array CGH together with a microarray expression analysis of ovarian cancer cell lines to provide evidence, consistent with data implicating ADRM1 as a 20q13 amplification target.
[pagebreak]
"Array CGH allowed us to identify the ovarian cell lines with and without gains and amplifications, and search through the genes that follow the expected pattern of over-expression in gained or amplified samples, compared to non-amplified samples. So using this technology greatly facilitates narrowing down candidate genes driving the amplicon," Fejzo says.
She adds that array CGH will continue to enable new cancer research discoveries for some time, as there are still many regions of copy abnormality where target genes have not yet been discovered.
Although Fejzo and her team analyzed 20q13, there are many additional regions that could be playing a role in tumorigenesis and could potentially be targets for designing novel therapeutics. "The advantage of array CGH is that, while we focused only on 20q13, we have the data on the entire genome and have the ability to do the same analysis on all these additional targets," she says. "With time, array CGH will allow elucidation of these genes and lead to personalized therapeutics based on the copy-number variation in each individual tumor — that is the future of personalized cancer therapy that is already starting in some cases."
One of the challenges facing investigators studying cancer with array CGH lies in complications that can arise as a result of tumor heterogeneity. "It's really not the right platform if you are interested in copy-number heterogeneity because if part of the tumor has an alteration and part of the tumor doesn't, you're unlikely to have the sensitivity to detect that — you'll have many cells with normal copy number swamping out those with an alteration," says John Iafrate, an assistant professor at Harvard Medical School. "So you need to be careful when studying primary samples and review the slides to makes sure that there is sufficient tumor cellularity. You want a high percentage of tumor, otherwise you will lose your signal to normal cells. They'll just swamp out the true findings in the tumor."
Iafrate and his colleagues published a paper in May in PLoS One that demonstrated the effectiveness of using array CGH to study recurrent chromosomal copy-number variants in sporadic chordoma — a rare, slow-growing tumor thought to form from the remnants of the notochord. Using an array CGH platform, the team performed copy-number analysis of 21 sporadic chordomas and found that large copy-number losses involving a number of chromosomes were more common than copy-number gains. The study also reported that deficiencies in -CDKN2A and PTEN expression might play key roles in chordoma pathogenesis.
"I think the days of non-array-based CGH are probably over; array CGH is becoming fairly standard now," Iafrate says. "We analyzed about 20 or so tumors from our bank using Agilent's 244K platform, and it goes without saying that that number of probes allows you to look very carefully within single exons. It has such significant coverage that we have fairly confident calls across the entire genome, which is pretty impressive, and the platform also has a good signal-to-noise ratio."
Iafrate says that at this point, it's too early to tell whether next-generation sequencing or array CGH will ultimately provide a more sensitive method for standard tumor analysis as there is just not enough data yet. And while algorithms for extracting copy-number data from next-generation sequencing exist, they are not widely distributed at this point, so the question remains open.
[pagebreak]
"One thing that is becoming clear is that we need to know as much as possible about every tumor, and array CGH allows you to get a whole-genome snapshot very quickly to see if there are any potential interesting targets," Iafrate says. "So we're currently exploring array CGH as a standard approach in tumor analysis, but the question will be, in the next few years, which is better — array CGH or next-gen sequencing for copy-number assessment? It's unclear right now."
Combined approach
In June, researchers at Uppsala University in Sweden and the Hospital for Sick Children in Toronto published a large-scale study in Nature Biotechnology that tested 11 different microarray platforms from three categories — SNP arrays, arrays containing both SNP and CNV probes, and CGH arrays. The CGH platforms evaluated in the study were Agilent Technologies' one- and two-sample 244K arrays and Roche NimbleGen's 720K and 2.1M CGH arrays. According to the team's findings, there is a less than 50 percent reproducibility rate for results when analyzing the same data across the platforms tested. In addition, the study found that the reproducibility of replicate samples is less than 70 percent, and, for variants in complex regions, the reproducibility was also lower than expected.
Although the authors concluded that the ideal strategy for CNV discovery was to use the newer generation of multi-probe, they were careful to emphasize that researchers should not necessarily choose one platform over another, such as CGH over SNP arrays.
The study did not, however, evaluate some of the emerging platforms that combine probes for CNV and SNP content, including Agilent's SurePrint CGH+SNP arrays. Roche NimbleGen also plans to release a similar platform later this year with higher-multiplex, high-density arrays for CGH applications.
Recently, a team of researchers demonstrated how array CGH can be integrated into a multi-pronged approach to study cancer. Jeremy Squire, director of translational research at the National Cancer Institute of Canada, reported in the journal Cytogenetic and Genome Research a proof-of-principle approach to elucidate the etiology and structural complexity of focal MYCN amplicons in human neuronal cancer. The team generated an integrated cytogenetic map of the MYCN amplicon using high-resolution array CGH — combined with spectral karyotyping, multi-color banding, and fluorescence in situ hybridization — in four neuronal tumor cell lines. Using this method, the group observed evidence of complex -intra- and inter-chromosomal events that could provide clues about the genomic underpinnings responsible for MYCN amplification.
"The information provided by the arrays was a little too complicated to interpret by itself. We thought we understood what was going on, but we had to see the cytogenetic entities that led to the amplification to reconstruct the geology of how we think things happened as a cell line involved these amplifications of this oncogene," Squire says. "We couldn't have done it with just array CGH and we couldn't have done it with just cytogenetics. It was really having the two together that made it work."
[pagebreak]
Despite its advantages, in some cases array CGH produces data on what appear to be copy-number gains, but what turn out to be insertions. Blake Ballif, director of research and development at Signature Genomics, published a study in March in Genome Research that described the importance of combining FISH visualization with array CGH to determine the nature and origin of copy-number changes. Ballif and his team reported 71 cases of unbalanced insertions identified by array CGH and FISH out of 4,909 cases that were referred to their lab for array CGH analysis, and found to have copy-number abnormalities. In two patients, the researchers found cryptic, submicroscopic duplications at or near the insertion sites, which made clinical interpretation of these insertions difficult. Using array CGH together with FISH and linear amplification, they were able to identify a 126-kilobase duplicated region from chromosome 19p13.3 inserted into MECP2 at Xq28 in one of the patients displaying symptoms of Rett syndrome.
"In our study, it turned out that, of the duplications we identified by array CGH, not quite 3 percent of those ended up being unbalanced insertions after we got done with some FISH experiments," Ballif says. "That was one of the real surprises to us, that that was so frequent, because a lot of the time we see copy-number gains in microarray data, and without doing FISH experiments you can't determine whether or not it's a tandem duplication event or if the duplicated segment is inserted somewhere else in the genome. That turned out to be more common than we anticipated."
At Baylor College of Medicine, Megan Landsverk is using array CGH in the institute's clinical lab for prenatal diagnosis of metabolomic disorders as well as for research. In March, Landsverk and her colleagues published a paper in Molecular Genetics and Metabolism describing their study of three families with known mutations causing metabolic disorders. While standard Sanger sequencing failed to detect these mutations, the detection of genomic deletions by mitochondrial/metabolic array CGH, called MitoMet, was crucial in providing a prenatal diagnosis.
"Next-generation sequencing technology has been touted as being able to pick up duplications and deletions, but it's not going to pick up really large things," Landsverk says. "If you are looking at anything over 500 base pairs, you're still going to need to use things like array CGH to find copy-number changes that next-generation sequencing can't pick up. Eventually they are going to come hand-in-hand."
For Landsverk, the key development to look for in emerging array CGH platforms is coverage density. "It's really about how many points can you put in one area, and as these arrays have developed, you've been able to add more oligonucleotides," she says. "So we're picking up things now on the arrays that we probably would not have picked up five years ago, just because the coverage wasn't as dense."
[pagebreak]
Considerations
When it comes to the adoption of array CGH in the cytogenetics community, the real hurdle is not always cost, but the challenge of teaching old dogs new tricks. "The classical cytogeneticists have found it a little harder because they have so much training in cell culture-based methods, but people who come from more of a molecular biology training [background] are very comfortable with array CGH," the National Cancer Institute of Canada's Squire says. "So we were fortunate in our group to have a mix of people who have an expertise at both levels, which is why we could do the paper. But it's true that sometimes in labs, it's hard to get the right mix of people familiar with arrays and cytogenetics."
While current array CGH platforms offer a level of resolution far surpassing that of traditional karyotyping, Harvard's Iafrate says potential adopters should become familiar with the structure of the human genome — including copy-number variation — before placing an order for the latest platform. "The major limitation for this [array CGH approach] is understanding benign copy-number variation. When you're looking at tumors you run into these copy-number polymorphisms, so you have to know to ignore them," Iafrate says. "If you're not experienced at all with the structure of the human genome, it's not something you can just jump into."
Many array CGH platforms do include software with built-in polymorphism databases for sorting out benign variations, making it a bit easier for researchers to adopt this technology. "There are people actively trying to adopt this technology in postnatal and prenatal research. I think there are fewer barriers than there were in the past from a technology standpoint," Signature Genomics' Ballif says. "There are some startup costs and a technology learning curve that people have to deal with, but I don't think that's a bigger hurdle than it has been in the past."