Background
The identification of chromosomal aberrations is critical to the appropriate evaluation of hematological neoplasms for diagnostic classification and risk stratification as defined by the World Health Organization (WHO) Classification of Haematolymphoid Tumours.1 Additionally, many professional society and expert consensus guidelines endorse testing for various SVs and CNAs for many hematopoietic malignancies. For example, National Comprehensive Cancer Network® (NCCN) Guidelines recommend cytogenetics by karyotyping for myelodysplastic syndromes (MDS)2, FISH for plasma cell myeloma (PCM)3, and both modalities for acute myeloid leukemia (AML)4 and chronic lymphocytic leukemia / small lymphocytic lymphoma (CLL/SLL).5 The European Leukemia Network (ELN) has issued similar position statements for AML and ALL.6,7
Traditional methodologies
CBA, or conventional karyotyping, has been routinely performed for over four decades.8 The chief advantage of this approach is the ability to discriminate subclonal architecture based upon microscopic visualization at the single-cell level. However, there are several drawbacks. Cell culture is required, which begets an extended turnaround time (TAT) of over a week yet may still yield elevated specimen adequacy failure rates in less mitotically active conditions. Despite some advances in technology (e.g., automated slide scanning), this technique remains largely manual and low-throughput, consuming time and resources. Result quality may also be operator-dependent, with skill generally increasing directly proportionate to experience. Finally, a fundamental disadvantage is the low resolution of ~5-10 Mb. Results may be incompletely characterized, with “marker” chromosomes of uncertain genetic content, and there is insensitivity for “cryptic” alterations of known clinical importance (e.g., NUP98 and MECOM rearrangements, inv(16) with CBFB-MYH11).
FISH utilizes DNA probes coupled to fluorophores to interrogate specific pre-selected alterations.8 TAT can be rapid if expedited on an emergency basis (within 24-36 hours) but is usually in the range of a few days. The resolution (~70 kb – 1 Mb) is approximately 10 times greater than CBA. The main disadvantages of FISH are that the targeted nature requires that the analyte of interest be known in advance and therefore dedicated reagents be available for those loci. While a limited number of markers can potentially be multiplexed, panels are generally required since there are now so many targets needed for complete evaluation. There can be additional limitations based upon probe design; for instance, break-apart probes do not identify the partner locus.
CMA technology, including array-based comparative genomic hybridization (aCGH)9 and single nucleotide polymorphism (SNP) arrays, is sometimes utilized to identify CNAs. The resolution is comparable to the lower bound of FISH at ~50-200 kb, and more automated procedures allow for higher throughput. However, a major shortcoming is the restriction to alterations with net gain or loss of genetic material; balanced translocations and inversions cannot be detected.10
NGS panels are most often applied to inspect for single nucleotide variants (SNVs) and small insertion/deletions (indels). NGS panels are also capable of identifying some CNAs and SVs but have several notable limitations.11 CNA detection is limited to DNA-based NGS, as RNA assays can only quantify relative expression levels, which are influenced by more complex factors than gene copy number. There are also typically size limitations that hinder accurate characterization of large-scale changes (such as full-arm or full-chromosome gains and losses).
SVs can also be quite challenging by short-read NGS. Targeted assays require that at least one partner gene be known at the time of assay design, and data analysis on short-read sequencing platforms may be complicated by pseudogene interference. DNA-based tests may require prohibitively extensive primer tiling across low-complexity intronic regions, while RNA-based assays require that a hybrid fusion transcript be expressed. For example, chromosomal rearrangements involving the IGH gene are highly prevalent in lymphoid disorders, yet extremely difficult to detect by NGS.12 Breakpoints are often intronic and highly variable, making DNA-based methods insensitive; further, the mechanism of action is placement of an oncogene under control of the active IGH promoter, resulting in increased expression rather than a chimeric RNA product. Similarly, oncogenic abnormalities involving the GATA2 and MECOM loci on chromosome 3q often alter expression levels without a change in mRNA sequence.13 Validation of appropriate thresholds for defining aberrant RNA expression is particularly challenging in the setting of variable neoplastic cell content, which may even be undefinable in some conditions such as MDS.
Recent and emerging technologies
To address the various shortcomings of current workflows, efforts have been made to improve CNA and SV detection while also streamlining laboratory operations.
Some of the pitfalls of targeted NGS panels regarding CNAs and SVs can be overcome by whole genome sequencing (WGS), which is well-established in the realm of constitutional genetics as being a single alternative to the combination of whole exome sequencing (WES) and CMA.14 In a study of 263 patients with myeloid neoplasms, WGS detected all 40 recurrent translocations and 91 copy-number alterations that had been identified by cytogenetic analysis, and also provided new genetic information in almost a quarter of patients, which changed the risk category for 16%.15 However, the significantly lower depth of coverage translates to decreased sensitivity for sequence variants, which are critical alterations to interrogate (i.e. by NGS); notably, these are not detectable by CBA, FISH or CMA. Limits of detection (LODs) for WGS-based tests are often double or more those of their targeted NGS counterparts (e.g., variant allele fractions, VAFs, of 10% vs. ≤5% for SNVs and 15-20% vs. 5-10% for Indels, respectively). Follow-up studies from the same institution utilizing long-read sequencing showed excellent correlation for CNAs and SVs with short-read WGS, but decreased accuracy for SNVs (96%, with 91% precision) and relatively poor performance for Indels (66%, with 42% precision).16 Nevertheless, long read WGS did improve upon reclassifying interchromosomal SVs called by standard WGS as intronic insertions near repetitive elements.16
Unlike any methodology previously discussed, optical genome mapping (OGM) utilizes total isolated ultra-high molecular weight (UHMW) DNA, which allows for spanning of repetitive regions and other regions that are difficult to map by shorter read methods. Nucleotide motifs are enzymatically labeled with fluorescent tags, and DNA is digitally imaged as it is linearized in nanochannel chip arrays; the resulting patterns are constructed into profiles, and CNVs and SVs are identified by bioinformatic comparison to a human refence genome database.17 OGM does not rely on tissue culture, limiting the possibility of culture artifacts, nor on PCR, preventing false overrepresentation due to amplification bias.18 Testing can be completed in as few as 4-6 days.19
Studies have reported robust performance of OGM across a spectrum of myeloid and lymphoid disorders and sample types, such as peripheral blood, bone marrow aspirate, CD-138 isolated cells, and lymph node cell suspensions.20,21 A wide range of CNAs and SVs have been successfully detected, including not only standard translocations and insertions/deletions, but also unusual karyotypic findings like ring chromosomes, isochromosomes, and markers.20 Additionally, OGM recognizes chromoanagenesis (chromothripsis, chromoanasynthesis, and chromoplexy), an indicator of severe of chromosomal damage.22,23 Chromoanagenesis is frequently associated with highly complex karyotypes and extensive clonal heterogeneity (known poor prognostic factors), as well as treatment refractoriness in AML.24 This phenomenon is extremely difficult, if not impossible, to detect by some currently used methods (e.g., CBA, FISH).24
The resolution of OGM is dependent upon depth of coverage and type of pipeline assembly (e.g., a de novo assembly has lower sensitivity compared to a rare variant pipeline, but can detect SVs smaller in size), but lower limits of ~500 bp – 5 kb can typically be achieved for insertions and deletions.19 While limits of detection can differ by the type of alteration under consideration, variant allele frequencies of 5-10% (theoretically translating to ~10-20% neoplastic cell content assuming heterozygosity in all tumor cells) are attainable, demonstrating comparable or superior sensitivity to other available testing methodologies.20 However, molecule alignment can be unreliable in regions concentrated around centromeres and telomeres, and while OGM can identify hyper- and hypodiploidy, it cannot dependably distinguish full or near-full polyploidy (e.g., triploidy, tetraploidy, etc.).18,19
A single-center study of 101 consecutive newly-diagnosed MDS patients23 demonstrated that OGM found all clonal abnormalities seen on CBA in 97% of patients. In addition, OGM identified 224 clinically meaningful findings not recognized by CBA (as well as ~50 calls of copy neutral loss of heterozygosity, CN-LOH, not detectable by CBA). OGM results changed the comprehensive cytogenetic scoring system (CCSS) and revised international prognostic scoring system (R-IPSS) risk-groups in 21% and 17% of patients, respectively, though the absolute improvements in prediction of prognosis were not statistically different than those obtained by CBA. However, according to a multivariate analysis, CCSS by OGM independently predicted survival but not CCSS by CBA.23
Investigators from the same institution subsequently described 159 AML patients (103 newly-diagnosed and 56 relapsed/refractory).22 OGM exhibited >99% sensitivity for detecting abnormalities identified by CBA and FISH when the clone represented ≥20% of cells; as would be expected, the few misses contained 3 instances of tetraploidy. OGM refined 17 fusion events called by CBA/FISH to name the gene(s) involved and reported additional findings in 59 patients including 11 critical MECOM, NUP98, and KMT2A rearrangements. Of significance, OGM also recognized KMT2A partial tandem duplications (PTDs), which cannot be detected by CBA or FISH and are most frequently interrogated by NGS. The authors noted that OGM results would have altered AML classification, risk stratification, and/or clinical trial eligibility in 24 patients (15%).22
OGM testing in a multi-center cohort of 100 AML patients25 demonstrated 98.4% concordance with sentinel events observed by karyotype at ≥5% allelic fraction and 100% concordance with FISH abnormalities seen at >10% allele frequency; sensitivity fell to 90.1% when CBA findings represented <5% allelic fraction. The authors acknowledge that these limits of detection suggest that OGM may not be suitable for detection of low-level clones, precluding use in the MRD setting. OGM also identified alterations not found by routine testing, but confirmed by orthogonal testing, in 13% of these patients. These additional findings would have changed ELN-2022 classification in five cases, and allowed for eight patients to enroll in clinical trials.25
A Belgian study of diagnostic samples from ALL patients26 investigated performance of OGM vs. routine testing. For 29 B-ALL patients, clonal abnormalities were identified by CBA in 79% and by FISH in 72% (with 93% by combined CBA + FISH) compared to 100% by OGM. Similarly, CBA combined with standard FISH identified at least one clonal abnormality in 8/12=67% of T-ALL patients whereas OGM had finding(s) in all of them. OGM failed to distinguish some subclonal aberrations (of no clinical relevance) seen by CMA/FISH but detected others at equivalent or lower frequencies. Importantly, disease-defining abnormalities were detected in 32 cases by OGM compared to only 23 by the standard testing pathway.26
In a comparison of OGM to FISH in 20 PCM patients27 with a minimum of 10% plasma cells following CD-138 enrichment, there were 100%, 92.5% and 95% accuracies for translocations, deletions, and gains, respectively; the lower values for CNAs were attributed to non-diploidy. Notably, all five canonical IGH translocations, gain of odd-numbered chromosomes, and abnormalities of chromosome 13 were detected by OGM. Additionally, OGM revealed additional prognostic markers in six cases, a 30% increase in yield, though follow-up data were not presented.27
An informative karyotype is unavailable in a substantial proportion of patients with myelofibrosis due to inadequate metaphase mitoses for analysis. In an investigation of 21 myelofibrosis patients in Spain,28 all samples generated successful OGM results (all confirmed by FISH or CMA, except one due to lack of available sample) while only 52% had informative CBA results.
The International Consortium for OGM in Hematologic Malignancies has issued consensus guidance29,30 focusing on validation, quality control, and analysis and interpretation of variants. OGM has also been incorporated into the American College of Medical Genetics and Genomics (ACMG) Technical Laboratory Standards,31 albeit for solid tumors, in which the body of knowledge is less advanced.32 Performing laboratories have suggested that a single-workflow methodology would be more time-, labor-and cost-effective.27,29
Additional novel technologies are also in various stages of evidence accrual, such as genomic proximity mapping (GPM), an NGS-based approach to capture ultra-long-range contiguity information from conventional short-read sequencing as a function of high-throughput chromosome conformation capture (Hi-C) plus artificial intelligence (AI). The underlying premise is that sequences located closer on a chromosome are more likely to physically interact and be crosslinked, so pairwise frequency of crosslinked sequence interactions can be used to determine the structure of chromosomes, such that proximity ligation signal increases as the genomic distance between any two loci across the genome decreases.33 A major advantage of GPM is the compatibility with archival (formalin-fixed paraffin-embedded, FFPE) samples. The preservation of chromatin structure and ability to assess 3-dimentional genomic architecture also allows for analysis of epigenetics.34 Similar to OGM, GPM also struggles to identify polyploidy and low-level allele burden.34
A preliminary study of GPM in 48 AML samples indicated 100% concordance with SOC for variants with impact on ELN 2022 risk stratification; GPM identified 39 additional variants, including variants of known clinical impact, not observed by cytogenetics.35 Another group reported that GPM showed complete agreement with all aberrations seen by FISH in samples from 5 newly diagnosed PCM patients, and also found supplemental clinically relevant or known recurrent abnormalities in 4/5.36 A third laboratory performed GPM on a mixture of 18 myeloid and lymphoid neoplasms with a >95% concordance with CBA, FISH, CMA and RNA NGS. GPM successfully detected balanced (and unbalanced) chromosomal rearrangements and CN-LOH, and findings not noted by SOC improved the accuracy of disease classification.37