Clinical Implications of Novel Genomic Discoveries in Chronic Lymphocytic Leukemia

Gregory Lazarian; Romain Guièze; Catherine J Wu

doi:10.1200/JCO.2016.71.0822

. 2017 Feb 13;35(9):984–993. doi: 10.1200/JCO.2016.71.0822

Clinical Implications of Novel Genomic Discoveries in Chronic Lymphocytic Leukemia

Gregory Lazarian ¹, Romain Guièze ¹, Catherine J Wu ^1,^✉

PMCID: PMC5559883 PMID: 28297623

Abstract

Chronic lymphocytic leukemia (CLL) is a common B-cell malignancy with a remarkably heterogeneous course, ranging from indolent disease with no need for immediate therapy to rapidly progressive disease associated with therapeutic resistance. The recent US Food and Drug Administration approvals of novel targeted therapies such as inhibitors of B-cell receptor signaling and B-cell lymphoma 2 have opened up new opportunities in the clinical management of patients with CLL and heralded a new era in the clinical treatment of this disease. In parallel, the implementation of novel sequencing technologies has provided new insights into CLL complexity, identifying a growing list of putative drivers that underlie inter- and intratumor heterogeneities in CLL affecting disease progression and resistance. The identification of these novel genomic features that can indicate future drug resistance or guide therapeutic management is now becoming a major goal in CLL so that patients can best benefit from the increasingly diverse available therapies, as discussed herein.

Chronic lymphocytic leukemia (CLL) is characterized by the accumulation and proliferation of clonal B cells in the blood, marrow, and lymph nodes. The clinical course of CLL is remarkably variable among patients, with median overall survival (OS) ranging from less than 3 years, despite the use of effective combination chemotherapy regimens, to more than 10 years without need of therapeutic intervention.^1,2 Genetic features underlying this variability in clinical course have been long identified. Conventional karyotype banding and fluorescent in situ hybridization (FISH) analysis are the bases of the existing widely used hierarchic prognostic model.³ The recent advent of transformative next-generation sequencing (NGS) has uncovered many novel putative disease-driving somatic alterations and accurately quantified intrasample heterogeneity. With the recent US Food and Drug Administration approval of inhibitors targeting CLL pathways,^4-6 the identification of the connections between each therapy and potentially vulnerable genomically defined disease subpopulations has become a priority in CLL.

UNCOVERING SOMATIC ALTERATIONS AND GENOMIC COMPLEXITY IN CLL

Over the last decade, new genome-wide sequencing approaches have identified numerous somatic alterations associated with cancer. These studies have provided a wealth of fresh insights into the underlying mechanisms driving cancer.⁷ The study of CLL has particularly benefited from the availability of these transformative technologies. With its ease of tissue accessibility and indolent disease kinetics, enabling repeated sampling within the same individual, CLL has been an optimal setting for examining questions of tumor heterogeneity and clonal evolution.

Earlier genetic studies of CLL that focused on the detection of copy numbers confirmed the recurrence of key cytogenetic abnormalities previously identified by FISH.^3,8 When considering data from an aggregate of 1,590 cases of CLL worldwide^9-13 (Fig 1A), the most common alterations and their frequencies have been focal deletions of chromosomes 13q [del(13q); 55% to 60%], 17p (3.5% to 10%), and 11q (6% to 27%) and trisomy 12 [tri(12); 10% to 16%]. Of note, the minimal deleted regions of these deletions were identified to encompass important putative CLL drivers: ATM and BIRC3 within del(11q), TP53 within del(17p), and the microRNA 15a/16-1 encoded within an intron of DLEU2 in 13q23.¹⁴ Other cytogenetic alterations were found at lower incidence, including chromosome 2p gains [amp(2p); 2% to 7%] containing MYCN, REL, and BCL11A,^15,16 amp(8q) (2% to 4%) responsible for MYC amplification,⁹ and del(8p) [2% to 5%] and del(15q) encompassing TNFRSF10A/B¹⁷ and MGA,¹¹ respectively.

Fig 1. — Putative driver gene mutations and recurrent somatic copy number variations in chronic lymphocytic leukemia (CLL). (A) Frequency of somatic copy number alterations when combining cohorts from the Spanish International Cancer Genome Consortium (ICGC) trial¹² (n = 452), German CLL Study Group CLL8 trial¹¹ (n = 353), Scandinavian SCALE trial¹³ (n = 369), Ouillette et al¹⁰ study (n = 255), and Dana-Farber Cancer Institute (DFCI)/Brown et al⁹ study (n = 161), identified through single-nucleotide polymorphism array (except as identified by whole-exome sequencing [WES] for Spanish ICGC cohort). Also shown are the frequencies of somatic single-nucleotide variants (from DFCI/Broad Institute¹⁴ [n = 548] and from Spanish ICGC¹² [n = 452]) identified through WES. Only the events over 1% in frequency in the combined cohorts are reported. *IGLL5*, *MAP2K1*, and *SAMHD1* have been reported in DFCI cohort only, whereas *ZNF292*, *KLH6*, *SETD2*, and *PAX5* enhancer have been reported in Spanish ICGC cohort only. All remaining genes were consistent among cohorts. Mutations in the 3′UTR of *NOTCH1* were detected in four of 150 patient cases with whole-genome sequencing. (B) Summaries of the characteristics of frequently mutated genes in CLL. amp, amplification; ANK, ankyrin repeat; del, deletion; EGF, epidermal growth factor; ERK, extracellular signal-regulated kinase; FAT, FRAP-ATM-transformation/transcription domain associated protein; HD, heterodimerization domain; MAPK, mitogen-activated protein kinase; PD, programmed death; PEST, proline, glutamic acid, serine, and threonine; PI3K, phosphatidylinositol 3-kinase; SCALE, Scandinavian Lymphoma Etiology; TAD, transaction activation domain; TET, ten-eleven translocation.

Even more recently, NGS technologies such as whole-genome (WGS) and whole-exome sequencing (WES) have provided a more granular investigation of the genomic landscape of CLL.^18-24 Rather than revealing any universal genetic event accounting for all CLL cases, these technologies have demonstrated the diverse recurrent gene mutations associated with CLL and the high level of genetic heterogeneity among samples, consistent with the high degree of clinical variability characteristic of CLL. The development of robust algorithms and statistical modeling of these sequencing data have led to the uncovering of gene mutations likely positively selected and hence identified as putative CLL drivers.^25,26

Two recent efforts using WES and WGS have together reported on approximately 1,000 patient cases of CLL, identifying recurrent gene mutations even of low frequency by highlighting a total of 75 significantly mutated genes.^12,27 In aggregate, mutations in 28 genes were found, common to both studies (Fig 1A). Conversely, the frequencies of gene mutations were dependent on the composition of the investigated cohort; 16 genes were unique to the Dana-Farber Cancer Institute (DFCI)/Broad Institute study, with SF3B1 as the most frequently mutated gene, whereas 31 were only detected within the Spanish International Cancer Genome Consortium cohort, in which NOTCH1 was the top mutated gene. This latter study also showed that relevant mutations can affect the noncoding genome (ie, NOTCH1 3′UTR and PAX5 enhancer). Although these two studies are the largest cohorts characterized by NGS to date, it has been estimated that at least 2,000 patient cases would be required to achieve saturation in driver discovery.²⁶

Examination of the growing list of putative CLL drivers has implicated the involvement of several key pathways in CLL biology (Fig 1A). TP53 and ATM are well-known tumor suppressor genes that are commonly inactivated by gene mutations or chromosomal deletions in CLL [del(17p) and del(11q), respectively; Fig 1B]. Such observations have identified the DNA-damage response pathway as a crucial CLL node. The application of NGS to CLL has also unexpectedly uncovered an important role of RNA processing in CLL. As a striking example, SF3B1 is a commonly mutated gene and encodes a component of the splicesome, which orchestrates the removal of introns from precursor mRNA.^19,21 Recently, SF3B1 mutations were shown to cause alternative splicing (with preferential alteration in 3′ splice site selection)^28,29 and a complex of changes, including impairment of the DNA-damage response³⁰ and alteration in telomere biology.³¹ Other frequently mutated genes involved in RNA processing and splicing have been identified (XPO1, RPS15, DDX3X, ZNF292, MED12, and NXF1), supporting the importance of this cellular process to CLL. NOTCH is another key pathway because it can be affected by either gain-of-function mutations of NOTCH1 (2 basepair frameshift deletion),²³ mutation of NOTCH1 3′UTR,¹² alternative splicing (by mutated SF3B1) of a pathway regulator,³¹ or FBXW7 loss-of-function mutations.³² B-cell receptor (BCR) signaling and the B-cell transcriptional program can also be impaired by mutations in EGR2, BCOR, IRF4, and IKZF3. Chromatin maintenance (CHD2, BAZ2A, ZMYM3, ASXL1, and SETD2), the inflammatory pathway (BIRC3 MYD88, TRAF3, and SAMHD1), mitogen-activated protein kinase (MAPK) –extracellular signal-regulated kinase (ERK; BRAF, KRAS, and MAP2K1), and MYC-related signaling (MGA and PTPN11) are other relevant pathways affected by mutations.

With the increase in the number of cases of CLL worldwide characterized by WES or WGS, it has been increasingly feasible to examine the likelihood that diverse somatic alterations cooperate to contribute to the oncogenic phenotype. Indeed, in the DFCI/Broad Institute study, most patients (approximately 60%) carried more than one driver.²⁷ Several studies have detected recurrent patterns of co-occurrence, highlighting likely preferred interactions between putative drivers. Del(17p) has been consistently associated with TP53 mutation,³³ del(11q) with ATM and/or SF3B1 mutation,³⁴ and NOTCH1 mutation with tri(12).^35,36 Conversely, low co-occurrence between SF3B1 mutation and tri(12) or NOTCH1 mutation suggests redundancy in their functional activities, as recently suggested by the finding that a target of SF3B1 mutation is a splice variant that dysregulates NOTCH signaling.³¹

In a reanalysis of our reported DFCI/Broad Institute data,²⁷ we found most CLL samples to display a unique combination of genetic alterations not occurring in any other patient sample, suggesting that each leukemia embarks on an independent evolutionary path (Fig 2). This analysis also revealed that for approximately 5% of patients, del(13q) is the sole detectable genetic abnormality. This event alone seems to be sufficient to drive CLL, because deletion in mice of the region corresponding to the human 13q14 led to the development of CLL-like disease, although at low penetrance and latency. Tri(12) and mutations of CHD2 and SF3B1 are also found as sole abnormalities in patients and could be sufficient drivers for CLL as well. Two thirds of recurrent combinations involved at least two of the following drivers: del(13q), del(11q), and SF3B1 and/or ATM mutations. Finally, 8% of CLLs in this cohort did not carry any known driver.²⁷ In these cases, there may have been putative genetic drivers not yet discovered; other factors such as epigenetic deregulation or microenvironmental factors could have played a role in driving disease.

Fig 2. — A summary of the frequencies and most common coassociations of putative chronic lymphocytic leukemia (CLL) drivers, per analysis of the Dana-Farber Cancer Institute/Broad Institute cohort of 538 CLLs.²⁷

Such genomic complexity defies single gene–based approaches for understanding cancer biology. Definitive understanding of the exact function of novel drivers and how they cooperate will require studies in in vitro and in vivo models.^37,38 Emerging single-cell technologies will likely transform our ability to decipher genomic complexity in CLL and highlight new drivers that could be relevant to individual patients. An expectation, however, is that these private drivers will nonetheless affect common core CLL pathways.

UNDERSTANDING CLL LEUKEMOGENESIS THROUGH ITS PHYLOGENETIC RECONSTRUCTION

Identifying founding genomic lesions and establishing when they were acquired may facilitate better understanding of the natural history of a case of CLL and suggest points of intervention. Because each mutation essentially supplies a molecular barcode for NGS reads, clustering of reads with similar variant allele frequencies (corrected for local ploidy and tumor purity) has feasibly allowed for accurate quantification of intratumor heterogeneity and investigation of clonal architecture and disease phylogeny. These types of studies have revealed events that are preferentially clonal, consistent with earlier events, and others that are preferentially subclonal, consistent with later events. On the basis of this principle, several studies have consistently categorized del(13q), del(11q), tri(12), and MYD88 mutations as early lesions, suggesting their role as CLL initiators, and ATM, SF3B1, and TP53 mutations as later lesions.^24,27 Likewise, the application of machine learning–based approaches has supported the idea that CLL-associated lesions are temporally ordered in a specific fashion rather than being randomly accumulated, again revealing del(13q) and tri(12) as early and potentially initiating events leading to preferred evolutionary trajectories.³⁹

A further layer of complexity that likely contributes to CLL disease heterogeneity is the impact of cell of origin on acquisition of subsequent somatic alterations. The methylome of CLLs with IGHV-unmutated status is more consistent with that of naïve B cells, whereas the epigenetic state of CLLs with mutated IGHV is more similar to antigen-experienced B cells,^40,41 although recent work has suggested that CLL cells can become fixed across diverse stages of B-cell differentiation.⁴² The state of B-cell differentiation seems to be associated with a preferential acquisition of certain somatic mutations, with mutated IGHV having a narrow spectrum of drivers (ie, MYD88, CD79A/B, and TLR2) and unmutated IGHV having a broad spectrum of events.²⁷ CLL-associated genomic abnormalities have been found even in the hematopoietic progenitor cells of patients with CLL. Deep sequencing of CD34+CD19-sorted cells of patients with CLL has revealed the detection of mutations in NOTCH1, SF3B1, and BRAF.⁴³ These findings are in line with murine xenograft studies in which engrafted hematopoietic progenitor cells from patients with CLL could produce a mature CD5+-expressing clonal B-cell population with a phenotype of CLL.⁴⁴

IMPACT OF SOMATIC EVENTS IN CLL ON RESPONSE TO THERAPY

The treatment landscape of CLL is currently evolving considerably. For more than a decade, the combination of fludarabine, cyclophosphamide, and rituximab has been the conventional first-line regimen for fit patients, providing a high response rate (90%) and prolonged progression-free survival (PFS; median, 57 months).¹ Although patients treated with this combination almost uniformly experience relapse, a subset of those with mutated IGHV can particularly benefit from this approach, with long-term PFS.^45-47 Other common chemoimmunotherapy-based monotherapy or combination regimens have included purine analogs (bendamustine, pentostatin, and cladribine),^48-50 alkylating agents (chlorambucil and cyclophosphamide),^51,52 and anti-CD20 (rituximab, ofatumumab, and obinutuzumab)^53-55 or anti-CD52 antibodies (alemtuzumab).⁵⁶ In addition to these chemotherapy-based regimens, recent US Food and Drug Administration approvals of novel therapies targeting BCR signaling (ie, idelalisib, an inhibitor of phosphatidylinositol 3-kinase,⁵ and ibrutinib, the irreversible inhibitor of Bruton tyrosine kinase [BTK]⁴) and the B-cell lymphoma 2 signaling pathway (ie, venetoclax⁶) have opened up the possibility of specifically targeting crucial CLL pathways, with less toxicity. Despite high response rates in patients for whom cytotoxic drugs have failed, resistance to these therapies is increasingly emerging.⁵⁷ In this setting, genomic characterization can identify the sources of resistance and help guide subsequent therapeutic decisions.

Intratumoral Heterogeneity Fuels CLL Resistance

It is increasingly evident that intratumoral heterogeneity not only fuels clonal evolution and leukemic progression but also provides the seeds for the development of therapeutic resistance.^24,27 Although therapy itself can potentially incite mutational events and increase genomic diversity,⁵⁸ multiple longitudinal studies using WES characterization in CLL have supported a scenario in which the genetic capacity for resistance is already present in the pretreatment sample as pre-existing subclones (Fig 3).^24,47,59-62 Alternatively, leukemic tumor cells at relapse could also be progeny of dormant parental cells. Recent findings showing the presence of somatic alterations in early hematopoietic progenitors are in line with this idea.⁴³

Fig 3. — Schema of clonal evolution of chronic lymphocytic leukemia (CLL) in relation to exposure to therapy. From the start, sequential acquisition of somatic alterations in CLL leads to its genomic diversification. Exposure to chemoimmunotherapy as first-line treatment commonly selects *TP53*-disrupted subclones. Each therapy proposed at relapse may differentially shape the CLL architecture and would be associated with preferential alterations: *TP53* loss with chemotherapy-based treatment and *BTK* or *PLCG2* mutation (mut) or 8p deletion (del) with ibrutinib therapy. ? indicates other mutations. FCR, fludarabine, cyclophosphamide, and rituximab.

Characterizing the Treatment-Specific Genomic Landscape of CLL

In the setting of exposure to broadly cytotoxic agents, TP53 disruption has been clearly identified as the most crucial and independent factor of resistance. Indeed, multiple studies have shown that patients with TP53 loss experience poor response and worse outcome after chemotherapy treatment.⁶³ In NGS studies tracking the fate of subclonal populations in patient samples before and after chemoimmunotherapy, subclones with disrupted TP53 clearly undergo clonal expansion by the time of relapse (Fig 3).^24,27,64 Importantly, small subclones with TP53 mutations detected at diagnosis only by sensitive techniques are associated with the same adverse prognosis as macroscopic subclones and can indicate future fludarabine refractoriness.^65,66 Subclones with TP53 mutations are also associated with lower death rates during therapy compared with subclones with TP53 wild type, suggesting diminished sensitivity to therapy and higher growth rates during repopulation.⁶⁷ These mechanisms underlying clonal dynamics have not been yet observed with other drivers and highlight once again the leading role of TP53 mutations in providing fitness advantage to the clone.

SF3B1 and ATM mutations seem to have variable evolution, with distinct clones rising or falling over time, suggesting that they likely do not bring the same advantage as TP53 when considered individually. By contrast, targeted characterization of the relapsing CLL genome has shown that combinations of mutations involving TP53, ATM, and SF3B1 could act synergistically to provide resistance to immunochemotherapy.⁶⁸ Thus, rather than sole abnormalities, combinations of somatic alterations could drive chemotherapy resistance in CLL.

By contrast, resistance to the targeted pathway inhibitor ibrutinib has been attributed to mutations directly affecting its target (BTK) or its downstream effector (PLCγ2).⁵⁷ BTK mutations (at the C481S site) are located in the ibrutinib binding site, resulting in a protein that is only reversibly inhibited by ibrutinib.⁶⁹ PLCG2 mutations likewise provide gain of function and lead to activation of BCR signaling in a BTK-independent manner.⁷⁰ Resistance to ibrutinib has been also associated with marked clonal evolution.⁷¹ Aside from BTK or PLCG2 mutations, relapsing subclones have been shown to be progeny of parental del(8p) leukemic cells, present before therapy. Del(8p) was found to generate haploinsufficiency of the tumor necrosis factor–related apoptosis-inducing ligand (TRAIL) receptor, resulting in TRAIL insensitivity that could contribute to ibrutinib resistance (Fig 3). Although ibrutinib represents a major advance in the treatment of patients with TP53 disruption,⁷² those treated with ibrutinib in this group of patients could nonetheless still have an adverse prognosis.⁷³ However, complex karyotype has been shown to be an even stronger predictor than del(17p) in this setting.⁷⁴ Thus, distinct modes of therapy may shape the CLL architecture differently.

Novel Approaches to the Detection and Monitoring of Subclonal Populations

Given that the bulk of recurrent putative CLL drivers have been discovered in WES or WGS studies (down to 3% frequency within the population), targeted NGS approaches are increasingly attractive as a means of efficiently and more routinely sampling the mutational landscape of CLL in a cost-effective fashion.^68,75 Deep-targeted sequencing displays high sensitivity, with the ability to detect mutated alleles in down to one in 100 or 1,000 cells, but it is still prohibitively expensive for the detection of rarer events.

Single-cell technologies are emerging as another powerful tool to probe the genomic composition in heterogeneous cellular populations.⁷⁶ Single-cell WGS approaches can potentially provide a comprehensive snapshot of the subclonal composition of a population,^77,78 but this is not high throughput nor cost effective at the present time. Droplet digital polymerase chain reaction technology is an alternative method for analyzing single cells by compartmentalizing them using water-oil emulsion at high throughputs. Using this strategy, the presence of ibrutinib-resistant subclones was recently been quantified before treatment initiation, demonstrating that drug-resistant populations were present in small quantities even in advance of ibrutinib exposure⁷¹ and that capacity for drug resistance was already inherent within the patient sample.

DEVELOPING GENOMICS-BASED SCHEMA FOR PROGNOSTICATION

Over the past decade, clinical prognostication in CLL has relied on the Rai and Binet clinical staging systems,^2,79 knowledge of the mutational status of immunoglobulin variable region (which separates CLL into either mutated or unmutated IGHV groups, the latter with worse outcome^80,81), and FISH assessment of the most recurrent chromosomal aberrations in CLL. In a hierarchic classification scheme, del(17p) was found to confer the poorest survival, followed by del(11q), tri(12), normal karyotype, and then del(13q) as the sole abnormality.³

A key question in the field is whether the increasing array of discovered putative drivers holds prognostic relevance and could improve the accuracy of prognostication. A growing body of validation studies has evaluated the associations between genotype and parameters such as time to first treatment (TTFT), OS, therapeutic response, and PFS (Fig 4).

Fig 4. — Summary of prognosis impact of somatic mutations evaluated in retrospective studies or clinical trials in chronic lymphocytic leukemia (CLL). The medians of time to first treatment (TTFT), overall survival (OS), and progression-free survival (PFS) in subgroups of patients with CLL with mutations are indicated in months and compared with subgroups of patients with wild type (in parentheses). When medians were not available in studies, the value of the hazard ratio in multivariable analyses is reported in square brackets. The UK Leukaemia Research Fund (LRF) CLL4 trial⁸³ compared median months for the *ATM* biallelic inactivation subgroup with *ATM* wild type and *ATM* mutation subgroup. Data from the German CLL Study Group (GCLLSG) CLL3X trial⁸⁴ were inferred from Kaplan-Meier curves. ERIC, European Research Initiative on CLL; FILO, French Intergroup on CLL; IPI, International Prognostic Index; MLL, Munich Leukemia Laboratory; NA, not available; NCRN, National Cancer Research Network; NR, not reached; ORR, overall response rate; R/R, relapsed/refractory; SCALE, Scandinavian Lymphoma Etiology. *5-year OS rate (%). †10-year rate (%).

Despite variability in the design and size of these studies, several consistent findings can be gleaned. First, TP53 disruption has clearly emerged as a reliable factor conferring adverse TTFT, PFS, and OS after first-line therapy.^33,63,82-84 In the recently reported CLL International Prognostic Index study, which evaluated 3,472 patients treated in prospective first-line trials, mutation in TP53 contributed the greatest weight to the score.⁸⁵

Second, the impact of SF3B1 and NOTCH1 mutations has been variably reported, which likely points to the importance of therapeutic context and patient status in evaluating this effect. In aggregate, several clinical trial studies have highlighted the independent poor prognostic influence of SF3B1 mutations on PFS or OS.^{33,82,84,86-88} Although NOTCH1 mutations have independently conferred adverse OS in multiple studies, their impact on PFS has been inconsistent.^87-90 Interestingly, an analysis of the international CLL8 study reported a lack of benefit from the addition of rituximab in patients with NOTCH1 mutations.⁹¹

Fewer studies have evaluated the role of ATM mutations; those that have been conducted have suggested an association with shorter OS.⁹² More recent studies, however, have argued for a notable role of biallelic rather than monoallelic inactivation in significantly shorter PFS and OS in the first-line setting.⁹³

Impact of other recurrent mutations remains under investigation. KRAS and POT1 mutations have been associated with refractoriness and adverse outcome.⁹⁴ EGR2 mutations were observed to be associated with shorter OS.⁴³ Although linked with poor prognostic features, MED12 mutations do not affect OS.⁹⁷ WES of large CLL cohorts recently identified additional novel prognostic factors associated with shorter TTFT (BRAF, ZMYM3, and IRF4), OS (ASXL1), and PFS (RPS15 and SETD2).^12,27,96

Given the genomic complexity and co-occurrence in drivers in CLL, sophisticated models and external validation are now required to understand the relative prognostic value of the host of diverse genomic events. Hierarchic models have been widely used for classifying cytogenetic lesions.³ By focusing on OS, a scoring system integrating both recurrent chromosomal aberrations and gene mutations was proposed to define four CLL risk groups: high (TP53 and/or BIRC3 abnormalities), intermediate [NOTCH1 ± SF3B1 mutations and ± del(11q)], low [tri(12) or normal karyotype], and very low risk [del(13q) as sole alteration], where survival is similar to the general population.⁸³ Prognostic scoring integrating a larger number of somatic alterations and focusing on both PFS and OS remains to be performed for accurately determining prognosis. The respective impact of somatic alterations is also expected to drastically change with the use of targeted agents, the resistance mechanisms of which may differ from those of conventional chemotherapies. An alternative approach would be to assign patients to subgroups based on distinct nonoverlapping molecular features, as recently proposed for acute myeloid leukemia, although this is challenging for a disease as genetically heterogeneous as CLL.⁹⁷

Given the consistent poor prognosis associated with inactivating TP53 lesions, clinical TP53 mutation testing [by FISH for the chromosome 17p locus and TP53 sequencing in cases without del(17p)]⁹⁸ is now highly recommended before each therapeutic line. Indeed, chemotherapy-based approaches have proven unsatisfactory in the presence of TP53 mutations and are now no longer the therapy of choice for these patients. Conversely, allogeneic transplantation remains a valid strategy for patients with TP53 disruption, as does the use of novel agents such as ibrutinib, idelalisib plus rituximab, and venetoclax.^6,89,99-101 Beyond TP53, sequencing of SF3B1, NOTCH1, and ATM may provide additional prognostic information, although therapeutic context will play a major role in assigning their relative importance. For analysis of clinical samples, a common panel of genes to integrate in patients with CLL could include TP53, SF3B1, NOTCH1, and ATM, given their frequency and potential prognostic impact. In the setting of ibrutinib treatment, the monitoring of gene mutations associated with ibrutinib resistance (ie, BTK and PLCG2) can define the potential need for alternative therapies. Larger panels of genes may be informative, but their prognostic impact remains to be determined (MYD88, BIRC3, KRAS, POT1, EGR2, MED12, BRAF, ZMYM3, IRF4, ASXL1, RPS15, SAMHD1, and SETD2).

In conclusion, the application of advanced genomic technologies has rapidly uncovered new mechanisms underlying CLL biology, disease progression, and therapeutic resistance. To fulfill the promise of precision oncology, future studies will need to address the challenge of translating these findings into the routine management of CLL so that patients can maximally benefit from recent therapeutic advances.

AUTHOR CONTRIBUTIONS

Conception and design: All authors

Data analysis and interpretation: Romain Guièze, Catherine J. Wu

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST